WO2010112007A2 - Capturing motions with feedback - Google Patents

Abstract

In a method for capturing motions of living beings by way of a mobile capturing unit, wherein a) the living being wears at least one motion capturing device, the motion profile of the motion capturing device is captured during a defined time period, a motion capturing device is used which comprises a pickup designed as a motion-sensitive sensor, the sensor captures motions in space, such as a 3D acceleration sensor, a 3D magneto-sensitive sensor or a 3D gyroscope, b) the living being wears an electronic evaluation unit, the data captured by the sensor is transmitted to the electronic evaluation unit, where it is automatically analyzed according to specified parameters, a comparison is automatically conducted to see whether the results of the analysis are within or outside of a so-called permissibility corridors, and c) the living being wears a signal transmitter, and the electronic evaluation unit automatically transmits a signal-triggering pulse to the signal transmitter when a result of the analysis is outside of the permissibility corridor, it is proposed according to the invention for a captured parameter of the sequence of motions to run as a curve that is plotted as a function of the time, and a pattern of motions is captured based on the sequence of motions, and events that occur within said pattern of motions are detected with precision in terms of time in that in each case two analysis values are correlated.

Description

 "Recording movements with feedback"

Description:

The invention relates to a method for detecting movements in living beings.

In particular, the detection of movements is at

Provided for people and horses. In the following, many people are mentioned, and this is only representative and exemplary for the application of the present invention in other living beings.

From practice, in sports known methods in which - usually by means of video technology - a sporty game is recorded and an analysis of the players later manually done by the game is evaluated by time-lapse technique. For this, each player has to be evaluated individually with a high staff effort. In other known from practice, a separate video camera is used per player, these multiple cameras must always be tracked, depending on where the players are on the field.

In a technically similar embodiment, it is known in the medical field to analyze a running on a treadmill patient, the analysis either simultaneously, by on-treadmill professionals such. B. can be done by a doctor, or can be delayed by means of a video recording.

The known methods have in common that they cause very high costs and are labor-intensive, and that they can only be carried out by experts. Related to the Medical application is also particularly disadvantageous in that the analyzes only possible movements that makes the patient under special conditions, quasi laboratory conditions, namely while he knows one hand observed white and on the other hand on a non-ordinary underground such as the treadmill mentioned as an example moves.

For example, if a patient should avoid certain movements after a hip operation, he will usually seek to learn this by means of physiotherapy exercises.

In this case, a so-called admissibility corridor is specified, which determines, for example, for the various joints of the human being certain angular movements in one or more directions which should preferably not be exceeded or fallen short of. It can not be ruled out that the patient in everyday life due to fatigue in the course of the day, or depending on its daily shape, or depending on the length of a distance already traveled, the movements that should be in the admissibility corridor, performs with very different accuracy and the admissibility

Corridor leaves more frequently than under special training conditions, such as B. during the performed under supervision physiotherapeutic exercises.

The invention has for its object to provide a method which allows a low-cost human movement analysis in humans. Furthermore, the invention has for its object to provide a device which enables the implementation of the method.

This object is achieved by a method according to claim 1 and a movement data acquisition device according to claim 19.

The proposed method makes it possible to judge the movements of people without the presence of specialized personnel. This can be done by analyzing the movement data the person or, if appropriate, an object moved by him. In the case of disadvantageous movements that lie outside of a previously defined admissibility corridor, this is signaled to the human being, so that the same training effect as in the aforementioned physiognomic exercises can be achieved.

By suggesting that a detected parameter of the sequence of movements runs as a curve plotted over time, and a movement pattern is detected on the basis of the movement sequence, and events occurring within this movement pattern are precisely identified in time by relating in each case two analysis values to one another, one can Evaluation of the motion sequence directly in the detection device or another carried by the living thing device done so that in the shortest possible time the living organism can be given signals that can influence his movements. Instead of a much later analysis of movements, as z. B. after sports events is common, for example, by means of the so-called bio-feedback even during the movement of the further movement of the living being can be influenced. Within the scope of the present proposal, bio-feedback refers to the fact that a physiological parameter that is not or not sufficiently perceived by the living being is converted into a sensible signal, so that the living being can react consciously or unconsciously to it ,

For the delivery of signals harmful to humans

The following consideration can be taken into account and For example, in applications in which the motion sequence, which can not be clearly determined, the desired improvement in the holding and the way in which it is moved, is preferably a so-called "biofeedback". Method of use. In doing so, the person is mainly informed of an improvement or deterioration of the desired posture / movement. As a result, the body has the opportunity to unconsciously perform the improvement. The exact knowledge of how to intervene in the complex sequence of movements is not absolutely necessary.

Motion-sensitive sensors are used as transducers, so that even without external reference points, the mobile detection unit can be used immediately anywhere (indoor and outdoor) without setup or calibration. The detection devices are self-sufficient in terms of power supply, so that no power connection during the detection of the movements is needed. The person - for example, the mentioned patient after a hip operation - can thus move freely according to his normal everyday life, so that movements can be detected as putting on a seat, climbing stairs, or running on different types of ground.

The proposed method is economically advantageous: firstly, for the detection, evaluation and possible correction of the movements no specialist staff must be present, secondly, the detection, evaluation and possible correction of movements is not limited to short periods in comparatively large distances, for example one hour per Day or week. Rather, training, as otherwise achievable by the physiotherapeutic exercises, can take place throughout the day by means of the mobile detection unit, ie, constantly when the person is moving, so that training progress can be achieved much more quickly and it can be avoided the human adopts unfavorable movements unconsciously. On the one hand, regular visits to specialist staff can serve to check the permissibility corridor and, if necessary, to redefine it, and on the other hand to document the progress of the training and to end a therapy after a corresponding success.

The handling of the devices is simple and without expertise to perform, since only the detection device to be turned on and must be carried by the living being. An evaluation of the data takes place fully automatically in the mobile registration unit. The computation time required for the analysis can be advantageously kept as short as possible, that only certain of the acquired data are automatically analyzed, for example, only the movement data, while actually also other data are recorded such. For example, human physiological data such as heart rate, skin moisture and the like, or environmental data such as humidity, air pressure or air temperature. All recorded data can be stored in a buffer and later transferred from the cache to a stationary computer at the mentioned specialist personnel, so that subsequently more complex evaluations of the recorded data can take place. This can be done either on the stationary computer itself, or from this connection can - for example, via the Internet - are built to a central server running on the expert program, and from which analysis results and, where appropriate, therapy proposals can be transmitted to the stationary computer.

The data can be transferred to a worldwide server via the Internet, with subsequent analysis. Subsequently, the results are transferred back to the above-mentioned transmission device, forwarded to another user, for example to another server. This other server may, for example, be operated by a medical center or the like. Long-term storage of the raw data and / or the analysis data on the server can advantageously be provided in order to be able to re-evaluate the stored raw data later, if appropriate from other analysis aspects, or to be able to make the stored results available again to the same or to another user ,

A general statistical analysis, such as For example, number of steps, (non-) regularity of steps, (A) symmetry of steps, and their time profile can be automatically generated based on the data stored on the server.

An analysis depending on the respective training, eg B. The maximum bounce, or the duration of that part of a Spaz ziergangs, at which the steps still exceed a certain Symmetriemaß, can also be automatically created based on the data stored on the server.

Likewise, the maximum as well as the minimum expression of different movements can be automatically created as an analysis result.

The results of the analysis can be used to create therapeutic measures.

The analysis results can be transmitted to a local user and / or to an external service provider, to a hospital operator or therapy organizer.

In the medical field, a "training concept" similar to that used in sport can also be used, for example in the context of rehabilitation measures, in which patients with limited physical flexibility perform exercises to improve their physical mobility be judged whether, for example, a range of movement of a first Due to the rehabilitation measures, the affected joint has become larger, but erroneous movement sequences can also be detected, so that, for example, protective postures and gentle movements that a patient has accustomed to, for example due to pain over an extended period of time, can be detected and corrected.

For example, after implantation of a new hip joint can be detected by the capture of biometric data of the patient's movement while walking, sitting down, getting up and in similar situations, so that even erroneous movement can be detected, which could for example lead to an element of the artificial Hip joint is dislocated.

If several patients carry out a joint gymnastic training, then a much more accurate and individualized analysis of the training progress can be performed subsequently or even at the same time on the basis of the acquired movement data than would be possible on the basis of the purely optical analysis for a single caregiver of this patient group. The advantage of cost-effective simultaneous care of several patients is combined with the advantage of an individual analysis of the patient's movements.

Another advantage of the proposed acquisition of biometric data is that not limited to the short period, the movements of patients can be analyzed, namely, if this patient with a doctor, physiotherapist or the like together, but that the data collection rather over a longer period can be done, for example, throughout the day, or over several days. Thus, for example, during a walk o. The like. The movement of the patient can be detected, so that at a later date, the data of

Qualified personnel can be analyzed and used for the appropriate treatment can be drawn for the treatment of the patient.

The analysis of the data may possibly also by an expert, so-called "expert system" in

Form of a computer program, which is installed, for example, on a server to which - possibly across country and state boundaries - possibly a variety of motion-capture devices from a variety of patients transmit their data, so that thereby the most economical and yet high quality Analysis of the biometric data and a correspondingly good care of the patients can be done. For example, the evaluation of the recorded biometric data to a doctor, physiotherapist o. The like. Be submitted, the

Evaluation has previously been done by a human or by the mentioned automated expert system, and this caregiver, so the doctor or physiotherapist, can give this patient suggestions in conversation with the patient or submit appropriate further therapy suggestions or terminate the therapy after a correspondingly advantageous course of treatment.

In a particularly advantageous embodiment of the present invention, it is provided to transmit signals in the form of acknowledgment signals or warning signals during certain movements which the analyzed person performs, for example a patient. Confirmation signals can be emitted, for example, if, during gymnastic exercises, for example, limbs have been moved over a wider range, for example beyond a further angle, than hitherto. In this way, the training or rehabilitating person is immediately signaled a training success. On the other hand, warning signals can be given off if incorrect movements are carried out, for example if a patient is still performing a movement sequence he has appropriated as a restraint or as a gentle movement. It may be advantageous to give up suitable movements and instead to move in another way, for example, to avoid overloading or unilateral stress on joints. By means of a warning signal thus a movement can be avoided, the z. B. after a hip implant the risk of dislocation of the implant salvage.

Thus, for example, a certain "movement corridor" can be defined and programmed into the movement detection device, so that the corresponding warning signal is emitted when leaving this movement corridor Such programming represents the mentioned contentual data transmission in the direction of the motion detection device.

A three-dimensional (3-D) representation of the recorded and cached movements is possible, so that the person can vividly show differences of permissible and impermissible movements and be explained by the qualified personnel.

The recording of the movement profiles by means of a small, light detection device that is worn by the living being or integrated into the clothing and / or in a man-held object.

The recording of motion-relevant data by means of motion-sensitive sensors such. B. 3-D acceleration sensors, 3-D magneto-sensitive sensors and / or 3-D

Gyroscopes.

Other movement-relevant data can be detected by means of air pressure sensors, temperature sensors and / or physiological sensors, such as, for example, As pulse, respiratory rate, skin moisture, breathing air or oxygen content in the blood. The acquisition device can compress the data. This speeds up the later data transmission to the stationary computer and also makes it possible to dimension the buffer in the mobile detection unit as small as possible.

A transfer of all data can advantageously be done automatically as soon as a connection to the stationary computer is established. This connection can be made by connecting the detection device to a transmission device by means of a plug connection, for example by means of an intermediate cable or by virtue of the detection device having a USB plug. The transmission device may be a PC with a USB port, or a special device that can be connected to multiple detection devices simultaneously, so that several detection devices of a human, namely, which are intended for multiple limbs of man, are connectable.

Long-term storage of the raw data and / or the analysis data on the stationary computer can advantageously be provided in order to be able to re-evaluate the stored raw data later, optionally under other analysis aspects, or to make the saved results available again.

It can be provided that the person does not carry the detection device directly on the body. In order to enable the most accurate detection of the movement of humans, can be advantageously provided to the movement of the

Do not interfere with people that motion-related data via a mobile, integrated into a piece of clothing or game device detection device with high temporal and spatial resolution and stored therein. In the simplest case, only a single measuring device is used, so that, for example, speed, acceleration and trajectory of the movement can be detected. Advantageously, however, it is provided that a person carries several measuring rates, so that the posture of the person and the movements of several of his limbs can be detected.

The acquisition device advantageously has sensors, a processor with its own firmware, a NandFlash as a buffer for the data, a real-time module with date recording, a power management system, a charger for a built-in battery and an interface for data transmission. Advantageously, the charger and the interface can be provided together, for example by means of USB

Functionality. An additional power connection is not necessary in this case.

In particular, the transmission device already mentioned above can advantageously be configured in such a way that it enables the simultaneous connection of a plurality of detection devices. Thus, not only the data of multiple detection devices can be recorded in a short time and transmitted to the stationary computer, but it can also be synchronized before the start of the automatic motion detection, the multiple detection devices, so that the internal real-time modules from a common start time based provide motion data , And so that due to this synchronized data acquisition, the individual movements later allow a precise analysis of complex movement patterns, such. B. movements of the trunk, head and limbs of a human can be accurately detected by the synchronization of the detection devices and analyzed together and possibly readjusted. The device can be charged both by the specialist personnel, eg by connection to the transmission device or to the stationary computer, as well as at home. The firmware automatically detects different hardware configurations and different versions. Up-to-date updates and upgrades can be loaded either automatically or manually with a USB connection through an integrated bootloader. The type of recording, individual channel frequencies, recording densities, etc. are adjustable before use, for example, controlled by the stationary computer or individually for each individual recording device.

Three-dimensional (3-D) acceleration sensors, 3-D magneto-sensitive sensors or 3-D gyroscopes are advantageously used as sensors, as well as FSR pressure sensors, air pressure sensors, temperature sensors and physiological sensors (for recording pulse, breath, skin moisture, respiratory air, Oxygen content in the blood, and / or myography). In the simplest version, only 3-D acceleration sensors and 3-D magneto-sensitive sensors are used. With these sensors it is possible to record a motion profile. The advantage is that the system can be used immediately for both indoor and outdoor applications because it requires no further infrastructure in the human environment.

In contrast to known devices, the mobile detection unit is a self-sufficient device that manages without information from the outside, such. B. of sensors that can also be configured as a camera. All position and movement data required for the analysis can be recorded by the device itself. The mobile detection unit can consist of several spatially separated, for example, distributed to humans attached components that interact functionally. An electrical power supply can from a central energy storage z. Wired to the individual NEN components be provided. A data transfer between the individual components can, for. B. using short-range radio such as the known Bluetooth radio standard.

The movement detection device can advantageously be very compact, so that it easily in clothes, shoes, protective equipment such. B. helmets or gloves, etc. can be integrated. Restrictions on installation, eg. B. because of an antenna to be considered, does not exist. In contrast to known devices that work with bending sensors, the sensors of a proposed detection device need not be mounted both above and below a joint.

Due to variations in the type and number of sensors, the system can easily be adapted by the manufacturer to the required resolution by producing appropriately equipped detection devices. Optionally, the user side in an expert mode, the activation or deactivation of individual sensors may be provided so that the amount of detected data can be adapted to the respective requirement and kept optimally low.

Slippage of the detection device, so variations in the

Attachment of the sensors on the body, can not be ruled out in practice. To compensate for this, it may be advantageous to convert the sensor data system into a body coordinate system. For this purpose, statistical methods can be used. These are based for example on the

Assuming that the main running direction is straight forward. Alternatively, before commissioning the detection device, or even in between, carried out a fast calibration that uses a short movement program, which consists for example of only two or three program steps such. B. 1. stand straight,

2. Move leg one step forward,

3. Lift the arm horizontally.

The data of all sensors are using a stochastic

State estimator, such as Kalman filters or particle filters, offset against each other to a position and position information. In the calculation boundary conditions can be considered such. As the location of attachment to the body / object and known body characteristics of the people / objects. These boundary conditions can also be extracted by statistical calculations from existing data records.

In one embodiment, the detection device is additionally equipped with a GPS receiver for adjusting the sensor data. The combination of 3-D sensor data and GPS position data is preferably also via a stochastic state estimator, e.g. Particulate filter. So z. For example, an average human step length can be calculated

All measurement data are stored together with time and date information in the internal buffer of the acquisition device. This time and date information comes from a realtime module that is integrated in the acquisition device. The synchronization and synchronization with other recording devices can advantageously be carried out automatically each time a connection to a PC with Internet access is made, for example by means of atomic-time-specific time data provided in a manner known per se by time servers and by means of which the stationary one Computer is automatically synchronized automatically. Each device additionally has a unique identification (ID number) assigned to a user. This allows the measurements of any number of devices to be synchronized with high precision in order to enable analyzes across them. In one embodiment, the detection device has no visible controls such. B. switch or button. In this way, it is impossible that the recording device can be switched off inadvertently or its settings can be changed accidentally. The recording of the movement data starts as soon as a certain, adjustable activity is exceeded. For example, the activity can be monitored by the acceleration sensor, which wakes the processor from a so-called sleep mode, or the

Recording can be started by certain movements, the z. B. by finger snapping or knocking against the housing as short-term and sudden accelerations are detected.

In one embodiment, the detection device is activated by means of a capacitive switch. The housing itself can serve as a capacitive switch.

In one embodiment, the detection device has a magneto-sensitive sensor, by means of which information for synchronization and / or to start the measurement are transmitted. This z. B. a battery-powered electromagnetic field emitting device used to start the detection devices. Several acquisition devices can be started at the same time. By means of coding of the magnetic signal, certain start parameters can also be transmitted.

Similarly, it may also be provided various

Status messages such. B. state of charge, still available data and / or electrical storage capacity, or triggering number of measurements by certain movements and visually and / or acoustically display with only a few display elements. So z. B. depending on the position in which Housing is on the table, different information is output.

In order to prevent a premature, unintentional switching off of the detection device, it can be provided that a shutdown does not take place by certain movements, impacts or the like but rather automatically, namely only when, for example, the above-mentioned monitored activity for a predetermined period has remained below a certain threshold.

In one embodiment, the sensor unit has a detection of whether and where it is associated with the garments. For this purpose, certain markings are attached to the garments as a signal generator, the one of

Apparent sensor element of the detection unit capacitive, optical, magnetic, via a Radio Frequency Identification Tag (RFID system) or mechanically detected by direct or indirect contact indirectly. Depending on the location at which the detection device in the

Clothing of humans is housed, thus a so-called position signal is generated. In this way, the recorded data can be automatically assigned to specific parts of the human body, while at the same time ensuring easy human handling. Thus, similar and similar-looking detection devices can be produced economically in large numbers, and the human being does not have to pay attention to which detection device is to be attached at which position of the clothing for detecting certain movements.

Advantageously, several detection devices are attached to the body. These units communicate with each other via a wired or wireless Body Area Network (BAN). The wireless variant can handle both RF signals, sound waves,

Light signals with or without light guide, on the body running Wei- len, use conductivity of the body in combination with electrostatic fields, magnetic fields (magnetic induction) or the like.

In one embodiment, the detection device is designed with a nonspecific shape narrow and elongated so that it can be accommodated anywhere as a separate element close to the body, for example, can be inserted into a stocking. Moreover, in this nonspecific shaping, without being specially adapted to a specific receptacle or mount, it can be arranged almost anywhere in textile pockets of garments, or in cavities of garment protective elements, or in cavities of objects carried by man, so that cost-effective detection devices can be produced in large numbers and can also be used inexpensively by the user for various applications.

In another embodiment, the detection device is arranged as an integrable element in a special protective part of the clothing. Thus, better protection of the detection device against mechanical damage is achieved than if the detection device were loosely stored or held in a bag or the like. The device advantageously has a plastic or metal housing, which in a special

Accommodation of clothing is housed.

Optionally, it may be provided that the detection device is inseparably connected to such a protective element. For example, the protective element itself can form the housing of the detection device and only have at least one connection element which makes possible the data transmission and charging of the energy storage device of the detection device. This can be z. B. be effected by the electronics of the detection device with the material of the z. B. plastic protection element is encapsulated. In one embodiment, the battery, z. As a Li-polymer battery, additionally protected by a special reinforcement of plastic or metal.

It is advantageously provided that the detection device meets the requirements of protection class IP65, that is protected accordingly against dust and moisture.

In one embodiment, the elongate detection device is additionally protected by a soft, damping material.

Advantageously, at least one light-emitting diode can be provided as a display, which is also visible in the ready state from the outside. It can be provided that the light-emitting diode is still visible from the outside even if the detection device is installed in a textile garment. It can, for example, signal the state of charge of the battery or give information about the recorded data.

The data transmission and the charging of the battery are each advantageously via a USB interface, wherein particularly advantageously the data transmission can also take place during the charging process. The USB connection can either be in the so-called built-in, ready-to-operate state or in the removed state, when the detection device is removed from the clothing or from the sports equipment.

In one embodiment, the USB port is only for charging the battery, and the data transfer to the cache is wireless, for example via a cellular network such as GPRS, UMTS, 3G, 3.5G, or 4G, or by other wireless standards such as wireless, Bluetooth, Zigbee or the like, wherein advantageously as widely used radio standard is used, so that a detection device can be used practically anywhere in the world. In a preferred embodiment, the raw data is compressed to reduce the amount of data prior to transmission and storage. For this purpose, a lossless compression is preferably used which achieves an average compression rate of at least the factor 5, in a more preferred embodiment at least the factor 7, in a most preferred embodiment at least the factor 9. If necessary, it can be provided that the data is further compressed by lossy compression. The compression can be done either before transfer from the acquisition device to the buffer, or before transfer from the buffer to the stationary computer. If the compression takes place before the transmission from the detection device to an external intermediate memory, or if the detection device itself has the intermediate memory, the compression can take place in particular directly within the detection device, so that the storage capacity required there for storing the measurement data must be as small as possible can.

Advantageously, a separate, otherwise required infrastructure for data transmission completely omitted: As soon as the detection device is connected via a USB port with the mentioned transmission device or said stationary computer, the data is transmitted to the stationary computer, and the time and date of Timebase in the acquisition device will be re-synchronized. This can be carried out particularly advantageously on the basis of the time information present in the stationary computer, so that all detection devices which are connected to this stationary computer are synchronized identically. In particular, if several such stationary computers are connected to a central server and in turn synchronized by this, worldwide analyzes are possible, for example influences of seasons,

Moon phases or the like on the movements. simultaneously The battery-powered detection device is charged when it is connected to the transmission device or the stationary computer.

In the mobile detection unit itself, an analysis of the acquired movement data is performed in order to inform the human on the basis of the emitted signals to adverse movements. An analysis that takes into account more or other measurement data in comparison, takes place in the shortest possible time automatically on the stationary computer or on the associated central server. In doing so, earlier measurement data and analysis results can also be taken into account.

In addition, specific factors can be identified, such as. B. for certain diseases or certain

Rehabilitation are typical. In one embodiment, the mean value and the standard deviation are analyzed over a specific time window and compared for conformity with given patterns or subjected to cluster analysis. Alternatively, factors derived from the raw data such as variance, energy or entropy, correlation between different axes or FFT coefficients, peaks in the raw data, or wavelet coefficients may also be used. The length of the time window is usually in the seconds range or in the near subsecond range and is preferably variable.

Preferably, the raw data is previously decomposed into DC (DC) and AC (AC) components via filters. For the recognition z. B. a hierarchical classification scheme can be used.

Data storage on the stationary computer or on a central server allows new analyzes to be carried out at any time. To make this possible, the - advantageously compressed - raw data are stored. The raw data Storage also allows a coherent overall view of the history of a person and thus also long-term analysis of his fitness state. Thus, the data are also suitable for anonymized analyzes of the fitness or health status of specific population groups.

The present proposal covers general medical and rehabilitation applications. It allows you to map the following four basic applications:

1. Diagnosis and compliance

2. Warning of harmful movements and biofeedback

3. adaptive training program

4. Recognizing emergency situations.

Medical application of sensors using the example of fall detection

The procedure comprises 3 different subtasks: the determination of the general fall risk (diagnosis and

compliance)

The acute detection of an increased fall risk (warning and biofeedback, training program) Alarm message after the fall (emergency situation)

The general fall risk is usually used to take general and longer-term measures, e.g. Structural measures, change of dwelling, support by a walking aid or the like, or it can be responded to by means of a behavioral change, eg. For example, "it may be better not to leave the house or apartment." The procedure can also be used in neurological conditions such as Parkinson's, Huntington's, and Cerebellar dysfunctions

The calculation of the acute fall risk has 2 tasks: 1. Increase the pre-warning time. This gives the wearer the opportunity to sit down, lie down or hold on time.

The actual fall is usually preceded by certain features: biomechanical z. As sudden persistence, solidification, fluctuations in standing z. B. by dizziness and the like. Physiologically z. As blood pressure change, sweating, the eyes are twisted. The feedback usually happens in the form of warning signals

2. Training measures to reduce the risk of falls: The risk of falling is calculated and reported back through biofeedback measures. An improvement or deterioration is presented in real time, with special exercises being carried out.

If the fall has occurred, a signal is automatically triggered and help is alerted.

For all 3 subtasks, the body movement is detected by the sensors and calculated together with the physiological data. The output for the risk profile is made on the PC in the form of diagrams, texts, etc. In addition, the risk is displayed on the device itself, e.g. by number / color / flashing frequency of 1 or more LEDs.

Sensor part:

For the motion detection, the information of at least one sensor which is attached to at least one body position, evaluated. The sensors are at least one acceleration sensor, one or more magnetosensors, one or more gyroscopes, one or more air pressure sensors, or a combination thereof. In addition, physiological data are recorded. In one application, sensors are attached to more than one location on the body. The sensors can be integrated in orthoses, or be designed outside of it as a separate unit.

The battery can advantageously be charged by movement, body temperature, light, etc., regardless of one

Power grid.

The sensor can be an inertial system for measuring 3D orientation / rotation. The buffer can be designed as a measuring memory for the data accumulated over days / weeks.

A sensor unit may consist of individual sensors for detection of 3D acceleration, 3D gyroscope, 3D magneto and extended by air pressure, temperature, FSR, strain gauges, or sensors for physiological factors. GPS and GPRS and MP3 can be integrated into the sensor unit.

Several air pressure sensors may be provided for differential measurement, z. B. to detect height movements when the person z. B. sets or lies down, or if he falls.

For acquiring physiological data, the sensor unit can be designed in the form of a keyword as follows:

• EMG electromyography

• Heart rate monitor for pulse rate, HRV

• skin resistance • blood pressure

• skin temperature,

• pupil movement by optical detection or dissipation of electr. Sum action potential in the header area

• Humidity

Procedure:

The isolated processes themselves may already be known. These are measurements in the laboratory under knew conditions and conditions, eg. B. 5 walking steps in the straight corridor or measurements on the treadmill o. Ä.

In the present proposal, the measured data are recorded under normal everyday situations. That can

Attitudes such. B. lying, sitting, standing, or movements such. Walking, running, jumping, climbing stairs, driving a lift, bus driving. Movements can be straight or curved, it comes to a sudden stop, one is triggered, etc. Each of these postures and movements influences and changes the sensor data and the physiological data. For an evaluation of the data, therefore, the knowledge of the movement is extremely important, for. B. Symmetry parameters / regularity can only be sensibly measured in straight forward gears. Fall parameters, however, are different, depending on how the initial position was, eg. For example, a sensory "free fall" from standing to sitting does not constitute an emergency situation, but a "free fall" from standing to lying is most likely a fall and thus an emergency situation.

Procedure:

1. Sensor data is acquired, calibrated, transmitted, stored

2. Coordinate transformation and filtering of the data

3. Recognition of everyday movement

4. Other methods depending on the type of everyday movement

5. Calculation of procedures 6. Display of risk on the PC and in the device

7. Biofeedback signals and warning signals

8. Adaptive training programs to improve

special procedures:

A) The automatic detection of everyday movements: Determination of feature vectors from AC / DC proportions, means, standard deviation, energy, entropy, correlation between axes, discrete FFT coefficients, wavelet coefficients, number and height of peaks, zero crossings in milliseconds / second range of one or more sensors on one or more sensors several body positions

Evaluation of the feature vectors via decision tables, nearest neighbor, decision tree, Naive Bayesian Classifier, HMM (Hidden Markow Model).

Pattern analysis, recognition of specific patterns in the sensor coordinates.

Recognition of various movements such. Stance phase / swing phase, step frequency, walking speed.

Grasping the movement trajectory: straight or not?

Movement direction of the COM by gyro or by magnetosensor.

RFID systems for recognizing places / objects.

An inventive method in the field of pattern recognition (pattern analysis) will be described in more detail below, in particular using the example of the application of the detection of a time of the footrest for timely step detection. In this case, parameters are recorded in time and plotted as a curve over time, the proposed method is explained in more detail below with reference to purely schematic Dartsellungen. In the illustrations shows 1a, b, c: stylized courses of the vertical acceleration signal in the heel piece of a walking human, recorded on the lower spine, FIG. 2: the time interval of the heel attachment points of 26 consecutive slow steps, at a measurement frequency of 100 Hz, so that 70 data points correspond to 0.7 seconds, FIG. 3 a shows an autocorrelation of the vertical acceleration signal of all 26 steps from FIG. 2, FIG.

3 a, in the form of the first six peaks, wherein the even-numbered peaks are comparisons of the side with the 1st, 2nd and 3rd following double steps of the same side of the human (ipsilateral), and FIG the odd-numbered peaks are comparisons of one page with the immediately following single step of the other side and the next and following pages of the other side of the human (contralateral), Fig. 4: the first six peaks of the autocorrelation of the vertical acceleration signal after time normalization of all 26 steps 200 values from step start to step start, FIG. 5: the first six peaks of the autocorrelation of the antero-posterior acceleration signal, FIG.

6: the first six peaks of the autocorrelation of the antero-posterior acceleration signal after time normalization of all 26 steps to 200 values from step start to step start, FIG. 7: the first six peaks of the autocorrelation of FIG

Yaw rate about the vertical axis (in the transverse plane),

Fig. 8: the first six peaks of the autocorrelation of

Rate of rotation about the vertical axis (in the transversal plane) after time normalization of all 26 Steps to 200 values from step start to step start,

Fig. 9: the first six peaks of the autocorrelation of

Rate of rotation around the anterior-posterior axis (in the frontal plane), and

Fig. 10: the first six peaks of the autocorrelation of

Rate of rotation around the anterior-posterior axis (in the frontal plane) after normalization of all 26 steps to 200 values from step start to step start.

For the pattern recognition, a descriptive algorithm is used, which allows to describe the pattern to be searched in a metalanguage. Previous methods use an amplitude coding based on multiple thresholds, in which the sequence of events takes place with or without time constraints (Eric Fimbel et al., 2003, Ivent Identification in Movement Recordings by Means of Qualitative Pattems, Neuroinformatics 1 (2003) p 239-257). For the present case of the exact time step recognition this is

Method not sufficient.

The problem can be seen in Fig. 1, which stylized shows three different peak shapes, denoted by a, b and c, of the vertical acceleration signal, which are measured with the footrest on the lower spine and typically occur within a series of measurements of several steps. Obviously, such a peak often contains several local maxima.

An algorithm that simply takes the highest maximum as a criterion would recognize this in c much later than in a and b. On the other hand, an algorithm that takes the temporally first local maximum within the peak would detect a peak much earlier in b than in a and c. Simple thresholds do not solve the problem because the peaks and their local extrema can be at very different amplitude levels.

These difficulties are usually circumvented by previously filtering the curves more or less strongly. This filtering has the disadvantage that it often changes both the exact time position of extreme values and the curve itself. To z. For example, to filter out the first maxiumum at curve 1b, relatively heavy filtering is needed.

Furthermore, filters are often compute-intensive and memory-intensive and often poorly real-time capable.

To solve this problem, we describe here a new form of peak analysis, called "Moving Triplet Analysis," implemented as a PeakFinder function, with each occurring local minimum or maximum being separately evaluated for validity When an extreme value occurs, an extremal triplet is examined in each case, at a maximum the triplet MinL-Max-MinR, from the previous left minium MinL, the maximum and the following right minium MinR, or vice versa, at a minimum the triplet MaxL-Min-MaxR, from the previous left maximum MaxL, the minimum and the following right maximum MaxR considered.

For the validity applies: The difference and / or the ratio of the central to the left and / or right extreme value must exceed a certain absolute or relative threshold value. This method allows the analysis and fault detection in one step, without the underlying curve previously z. B. must be greatly changed by mean value in their appearance. In addition, when processing from left to right from the central maximum and MinR for the following triplet MaxL and central minimum, so that ultimately only two

Values must be saved. In the case of failure to reach the The existing extremata for the left triplet side remain valid. This economy, as well as the ability to process the incoming data immediately, is very important for calculations in small, portable devices, especially if the calculation is to be performed while walking for the purpose of immediate biofeedback.

The method according to the invention is further characterized in that advantageous for the description of amplitude curves

Peaks are used whose properties relate to each other within the search string, also called relational peak analysis. A peak is a coherent, suprathreshold range of measured values, which is defined by various parameters. The position in the search string is determined by the time sequence of the peaks in the measurement channel. For z. As the timing of one or more maxima / minimum values, the start or end of a peak or the position of the center of gravity. For relative references to each other, qualitative measures such. B. the area, width,

Height of extreme values, but also descriptions of the shape of the peak (Gaussian, Kurtosis) or descriptions of an equivalent normal distribution with mean and standard deviation possible. The following example serves as an explanation:

channel_122: MaB2! Peak MiA4 minSteps: 1 0 1 maxSteps: 1 ++ 1

[qState4] peakarea> 2 0.6

The temporal succession of events is predetermined by the order within the search string (MaB2,

! Peak, MiA4), mean: channel_122: is the measurement channel to analyze minSteps is the minimum number of samples that the event must be valid. A 0 indicates that the event does not have to occur at all. maxSteps denotes the maximum number of samples that the event may be valid, a ++ indicates that the length is arbitrary.

MaB2 refers to the temporal event maximum of the analysis function Peakfinder B. The number 2 serves as the address for the relative reference.

IPeak means that in the defined time period no (!) Extreme value found by the peak finder may occur, ie according to MaB2 the next extreme is MiA4, but any number of samples may be later. MiA4 refers to the temporal event Minimum of the analysis function Peakfinder A. The number 4 serves as the address for the relative reference.

[qState4] returns a relative condition within the search string: the area of the peak with the relative address 4, ie MiA4, must be larger (>) than 60% (0.6) of the area of the peak with the relative address 2, ie MaB2 ,

It is also possible to process relative references between peaks of different channels. In addition, derivations can be considered for each channel, in particular the

1st, 2nd and 3rd derivative.

[PeakFinderA] negatePeak false; Data values must be greater than the value of the threshold. For negatePeak true, they must be less than the value of the threshold. Threshold 0; Value of the Threshold thr_slope_left_proz 0.6; minimum relative difference of the middle extremum to the left triplet partner in percent to a reference value. thr_slope_right_proz 0.6; minimum relative difference of the middle extremum to the right triplet partner in percent to a reference value. thr_slope_left_abs 50.0; minimal absolute difference of the middle extremum to the left triplet partner. thr_slope_right_abs 50.0; minimal absolute difference of the middle extremum to the right triplet partner. thr_slope_ratio 0.9; minimum required asymmetry of the differences to the left and right triplet partners.

[PeakFinderB] negatePeak true; Data points must be smaller than that

Threshold criterion

Threshold -5; Only data points smaller than -5 are examined thr_slope_left_proz 0.1; so. thr_slope_right_proz 0.1; so. thr_slope_left_abs 20.0; s.o thr_slope_right_abs 20.0; so. thr_slope_ratio 0.4; so.

When the PeakFinder function is called, it is still decided whether it should return all or only the first, the most extreme or the last of the local extrema.

For the task of always recognizing the step in FIG. 1 a, b and c at the apparently same point, the conditions in the peak finder function must be set such that the first local maximum in FIG. 1 b does not reach a validity step, because the thr_slope_right is not reached, and that only the first valid local maximum is returned as a result of the peak analysis.

This characteristic does not correspond in time exactly to the beginning of the footrest, but during the expiration of the footrest exactly recognizable time and occurs about 50ms after the beginning of the footrest on. For the most accurate detection of the beginning of the footrest so-called relational peak analysis is used, which uses additional peaks in the first and / or second derivative of the acceleration signal within this temporal reference frame. Another application of relational peak analysis is, for example, the detection of foot impact which, together with the footrest, divides the step into stance and swing phase. Another application of relational peak analysis is, for example, the detection of kick-off and jump-in-take, which together determine the flight time and from which the flight altitude can be calculated, an important parameter for the measurement of the speed force.

Another application of relational peak analysis is the

Analysis of the gait pattern of horses using the signals of the motion sensors (eg acceleration sensors, gyroscopes and magnetic field sensors) at a central measuring point (eg at the sternum or at the back.) In quadripods, the signals are transmitted through the footrest and repulsion of all 4 legs, which occur in different step patterns (walking, trotting, pass, booby, gallop) is a relational peak analysis of great advantage in the precise analysis, such as for detecting movement anomalies (limp, stress, etc.), since - If necessary with a lot of effort motion sensors are required at several extremities to allow a correspondingly accurate analysis.

B) The measurement of the symmetry of the movement:

By autocorrelation and / or correlation.

By calculation and comparison of gear-specific parameters right / left.

By autocorrelation / correlation of shear events (eg contact time) timed steps. The temporal normalization of the steps carried out before the autocorrelation calculation is advantageous if the analysis can not assume a constant stepping rate, as usually occurs only under laboratory conditions. For measurements in everyday life, variable stepping speeds are the norm. The following is an example of how the time normalization of the steps before applying the autocorrelation makes it possible to derive therefrom a parameter for the gait symmetry, which is independent of short-term variations in the step speed.

A measurement of a slow walking human became a sequence of 26 contiguous ones

Steps detected and extracted as a block. The measurement frequency was 100 Hz, so that measurement times were made in 10 ms distance. The time interval of the 26 steps, detected from the signal of the foot attachment in the vertical acceleration signal, averaged 69 measurement times with a standard deviation of

3 measuring points or 4.3% of the mean. This value is normal, but it is much higher in patients undergoing hip replacement, with an average of around 12%.

Fig. 2 shows the sequence of pitches, which can recognize no particularity and is quite normal. Nevertheless, the autocorrelation of the vertical acceleration, shown in Fig. 3a, shows that the height of the peaks drops very rapidly to a value of 0.4. This is because individual steps deviate from the normal time step length and thus the

Peaks in the autocorrelation calculation can no longer be brought into line, so the corresponding Kurvenabschnite so can not be superimposed exactly, if they are only moved by a fixed predetermined amount of time. The height of the peaks of the autocorrelation calculation becomes very high sensitive already influenced by small temporal variation in the peak patterns of the signal, in this case the steps.

In autocorrelation, the even-numbered peaks compare with the same side of a double step (ipsilateral), the odd-numbered peaks compare with the other side (contralateral). A measure of the gait asymmetry is the ratio of the height of the first peak to the second peak, that is, how much smaller the contralateral regularity with respect to the ipsilateral regularity of the step pattern is pronounced (Moe).

Nilssen and Helbostad 2004, Estimation of Gait Cycle Characteristics by trunk accelerometry. Journal of Biomechanics 37 (2004), 121-126). If the first peak is about the same as the second one, there is no asymmetry because the same amount of immunity occurs even with steps on the same side. The large drop, shown again enlarged in FIG. 3b for the first six peaks, and caused by the temporal variability in the steps, strongly disturbs this calculation.

The usual purpose of an autocorrelation calculation, to develop regularity in the measurement signal without its prior knowledge, is not impaired, as the clear expression of regular peaks in FIG. 3a shows whose distances correspond to the average frequency of the pattern regularity, in this case the step pattern.

It is an object of this invention to make the autocorrelation calculation useful for detecting gear asymmetries from variable-speed measurement data. This is made possible by first recognizing a characteristic detected in the steps, e.g. As the footrest in the vertical acceleration signal, and the measured values are interpolated between the times of these characteristics, so that after this time standardization all steps have the same length. 4, the first six peaks of the autocorrelation of the vertical acceleration signal are shown according to FIG. 3b, but now from the temporally normalized data after interpolation of all 26 steps to 200 measured values from footrest to footrest. The characteristic need not be timed exactly a known event, eg. B. the beginning of the footrest, correspond, but only at every possible event, such. B. footrest, be precisely timed recognizable. This is for example given by the above-described PeakFinder function by means of moving triplet analysis, which is a

Characteristic (first valid local maximum in Fig. 1) detects about 50 ms after the beginning of the footrest.

As a result of this normalization, the even-numbered (ipsilateral) peaks 2, 4 and 6 are at approximately the same level of 0.91, so that, with the temporal variability excluded, the curve from double step to double step is 91% similar. In comparison, the odd-numbered (contralateral) peaks are only slightly lower, so that no significant asymmetry can be seen in this signal.

The autocorrelation of the measurement series of the anterior-posterior acceleration channel of the same sequence of steps is shown in FIG. 5 (normal, unnormalized) and FIG. 6 (after time normalization) again for the first six peaks. A symmetry can clearly be seen in FIG. 6 in that the odd-numbered peaks, situated between 0.85 and 0.9, are significantly lower than the even-numbered peaks at just 0.95. In Fig. 5, this fact is obscured by the progressive decrease in the peak height and can only be recognized by the fact that this decrease is not uniform but stepped.

FIGS. 7 and 8 show the first six peaks of the autocorrelation of the measurement series of the rotation rates about the vertical axis (in the transverse plane) of the same sequence of steps, again without (FIG.

7) and with (Fig. 8) temporal normalization. Here were the here For better optical visibility, the autocorrelation absolute values are used because the odd-numbered peaks are negative because the yaw rates about the vertical axis are opposite for left and right steps. In the autocorrelation after time normalization (FIG. 8), an asymmetry can be seen from the different height of the odd-numbered peaks to the even-numbered peaks, this asymmetry not being visible in the normal, non-normalized autocorrelation calculation.

9 and 10 show the first six peaks of the autocorrelation of the measurement series of the rate of rotation about the anterior-posterior axis (in the frontal plane) of the same sequence of steps again without (FIG. 9) and with (FIG. 10) temporal normalization. Here, the absolute values of the autocorrelation were used here for better optical visibility, since the odd-numbered peaks are negative, because the rotation rates around the anterior-posterior axis for the left and right steps are in opposite directions. In the autocorrelation after time normalization (Figure 10), an asymmetry is seen from the different height of the odd-numbered peaks to the even-numbered peaks, which asymmetry is obscured in the normal, unnormalized autocorrelation calculation similar to Figure 5 by the progressive decrease in peak height and leaves its mark only in the staircase-shaped decrease.

Thus, a robust measure of asymmetry of step parameters in signals of acceleration and yaw rate sensors is obtained by temporally segmenting the steps based on a particular step feature. This is preferably done with the ones described in the previous section

Methods of the PeakFinder function with "moving triplet analysis" and relational peak analysis.These steps, which are now well defined in terms of time, are then normalized in time by interpolation before autocorrelation is calculated on these normalized step signals, which is robust against short-term variations in the temporal step and correlates the height of the odd-numbered autocorrelation peaks from time-normalized data with the height of the even-numbered autocorrelation peaks from time-normalized data and varies only slightly above the first six peaks.

In addition, the method according to the invention has the advantage that it is much less computationally and, above all, memory-intensive, since it is not necessary to calculate the entire autocorrelation curve, but only the value of the peaks themselves, the position of which is known in advance due to the temporal normalization , This saving, in the example shown about a factor of 70, is very important in calculations in small, portable devices, especially if the calculation is to be performed while walking for the purpose of immediate biofeedback.

A possible temporal asymmetry during walking is indeed lost in the normalized autocorrelation, but since the step times are known exactly, it can be examined exactly from these. Symmetry and regularity of the waveform (separate in each measurement channel) and the step times are examined separately using this method and thus allow a separate and more accurate analysis.

C) The assessment of the regularity of the movement:

Autocorrelation and / or correlation

by calculation and comparison of gear-specific parameters right / right or left / left

by autocorrelation / correlation of step-specific events (eg contact time) of temporally normalized steps, as described in the previous section on symmetry. Due to the temporal normalization, in steps with temporal variati- is defined as a robust measure of regularity by the even-numbered autocorrelation peaks after time normalization (Figures 4, 6, 8 and 10). Without this time standardization the even-numbered peaks of the autocorrelation are very inconsistent (FIGS. 3, 5, 7 and 9). The temporal regularity is, as already explained above, considered separately.

D) The detection of a fall risk:

Symmetry and regularity of movement, as described in the previous chapters on symmetry and regularity.

Variation width anterior / posterior or lateral / medial.

Gear speed / energy consumption

Special test procedure / test conditions z. Eyes, soft ground, getting up, getting up

Sit down

Frequency domain analysis: amplitude of harmonic oscillations

E) The detection of an overload

Duration and intensity of movement, of the joints

Recalculation of measured sensor data to other body parts / joints by inverse kinematics

F) The determination of statistical parameters for physicians and therapists z. Eg distance, speed, calorie consumption ... G) The transmission of data over the Internet

H) The change of physiological parameters

I) The automatic detection of a fall situation: This is described below in a separate chapter:

fall detection

The proposed method takes into account at least one of the following values derived from the sensors: intensity of the sensor signals

Position of the body parts that are connected to the sensor (s) in relation to gravity. This will be the

Low-pass filtered and averaged sensor signal used. Dynamics of the body parts operatively connected to the sensors.

The intensity of the sensor signals is used to

^• a shock = over-threshold value

^• an unbraked case = 1g of body center of gravity

^• Motionless = subliminal value, no change in the signals to detect.

The position to gravity is used to

^• the head posture = angle to gravity

^• the shoulder girdle parallel to the ground = angle to gravity • normal postures such as standing, sitting, lying and deviating postures to recognize = multiple sensors in different body positions such as shoulder, leg, trunk.

The dynamics of the sensor signals are used to Movements like walking, running = comparison stance phase / swing phase leg

Crawl = slow speed while lying, near the ground • Position changes such as getting up = integration or

Summation of the acceleration to speed and the distance covered the fall height = integration or summation of the acceleration to speed, distance covered.

The air pressure sensor detects a stay near the ground.

Fall detection is either only the intensity (impact, unbraked case, immobility), only the position, only the

Dynamics or a combination of the mentioned factors used.

In one application, a timely combination of head posture intensity values is used. A fall is given when either a shock, an unbraked fall or a large height difference of a body part with a not upright head posture coincides in a timely manner. In another application, an event is considered a fall when either a shock, an unbraked fall or a large height difference of a body part coincides with the non-parallel position of the shoulder strap to the ground in a timely manner.

In an application, a timely combination of intensity values with the position values "do not sit ^1" , "do not lie ^1" and "do not stand ^{1" are} used. A fall is given when either a shock, an unbraked fall or a large height difference of a body part does not coincide with any of the described postures or if an unusual posture is detected. In one application, a fall event is detected by detecting the 'crawl' behavior or at least one pending attempt ¹ event.

In another application, a near-ground location, z. B. mediated by the air pressure sensor, which exceeds a certain period of time, exploited to recognize.

In one application, an alarm signal can be issued wirelessly.

In one application, a tiered alarm may be issued with varying probability.

Biofeedback method for fall prevention: is fully integrated in orthosis is attached to the body

^■ via external device eg. Eg PC

Coupling with sensor part via wire or wireless • Mobile phone or PDA as output device

^• graded method, easy, medium, "free-wheeling" is adaptive, sets different sensitivities independently

^■ is at another site of action, eg. B. insole, toe, glove, shoulder, socks

^• as a single element or matrix

The biofeedback signal can be given as vibration, e.g. B. with adjustable amplitude

The biofeedback signal may be tactile, e.g. B. with pneumatic or hydraulic control. The energy for activation may be gained from body movement itself (eg, a rubber ball may be provided in the shoe, the joint movement actuating a pneumatic or hydraulic pump, or an electric generator may be provided). The tactile stimulus can be generated by using a parenthesis more or less pushes or a mandrel extends more or less, or a protective layer moves more or less back and releases a tactile element (eg a matrix with cover). The biofeedback signal can be given optically, e.g. B. 5 by a arranged in a pair of glasses lighting element such.

B. one or more, possibly different colored LEDs. An LCD display can be mirrored into the glasses through which the person can see through. One or more laser pointers project more or less

10 ße points on the floor, or on other surfaces such.

B. a wall surface.

The biofeedback signal can be given auditory, e.g. B. tones can be transmitted via Bluetooth to a receiver in the ear, z. B. to a conventional hearing aid.

15 The tones can be changed in intensity, frequency or spectrum to transmit different signals. Voice instructions can be given in plain text. The biofeedback signal can be given electrically, e.g.

For example, a stimulus current may be delivered via one or more electrodes. Electrode attachment and material may be as required. EMG.

Claims

claims: A method for detecting movements in living beings by means of a mobile detection unit, wherein a) the living entity carries at least one movement detection device, during a certain period of time the motion profile of the motion detection device is detected, a motion detection device is used, which one as movement sensitive Sensor has designed transducer, the sensor detects spatial movements, such as a 3-D acceleration sensor, a 3-D magnetosensitive sensor or a 3-D gyroscope, b) the living organism carries evaluation electronics, which detected by the sensor Data are transmitted to the evaluation electronics and there automatically analyzed according to predetermined parameters, a comparison is automatically carried out, whether the analysis results are inside or outside a so-called admissibility corridor, c) the living being carries a signal generator, and automatically sends a signal from the evaluation electronics. triggering impulse to the When an analysis result outside the admissibility corridor is transmitted signal is transmitted, characterized in that a detected parameter of the movement sequence as a time-applied curve, and a movement pattern is detected by the movement process, and occurring within this movement pattern events are accurately recognized in time by two analysis values are related to each other the. 2. The method according to claim 1, characterized in that for precise timing of the event, a so-called moving triplet analysis of the curve is performed, wherein either a local minimum of the curve, which is located between two local maxima, or a local maximum of the curve , which is located between two local minima, when the event is detected, when the triplet consisting of minimum and maximum or maximum and minimum minima reaches or exceeds predetermined thresholds. 3. The method according to claim 1 or 2, characterized in that a complex movement pattern is detected by means of the so-called relational peak analysis of the curve, wherein the peak is a contiguous, suprathreshold range of measured values, which is defined by specific parameters , and wherein the properties of at least two peaks relate to each other within a search string. 4. The method according to any one of the preceding claims, characterized in that symmetries and deviations of symmetries that occur in the detected movement sequence, are detected by means of a time-normalized autocorrelation. 5. The method according to any one of the preceding claims, characterized in that by means of several transducers, the motion individual body parts of the living being. 6. The method according to any one of the preceding claims, characterized in that the plurality of transducers used are synchronized before the data is transmitted to the evaluation electronics. 7. The method according to claim 6, characterized in that the transducers are synchronized before the start of the specific period of time during which the motion detection takes place. 8. The method according to any one of the preceding claims, characterized in that by means of physiological sensors further movement-relevant data of the living being are detected, such as the pulse rate, the skin moisture or the respiratory rate. 9. The method according to any one of the preceding claims, characterized in that by means of living beings worn sensors further motion-relevant, but independent of living organisms data are detected, such as the ambient temperature or the ambient air pressure. 10. The method according to any one of the preceding claims, characterized in that the data collected by the transducer and / or the analysis results are transmitted to a living carried by the buffer. 11. The method according to claim 10, characterized in that the data collected by the transducer and / or the analysis results are compressed before being transferred to the cache. 12. The method according to claim 10 or 11, characterized in that the data stored in the buffer data is transmitted to a stationary computer. 13. The method according to claim 12, characterized in that the present on the stationary computer, detected by the transducer data are analyzed according to other parameters than in the carried by living organisms evaluation electronics. 14. The method according to any one of the preceding claims, characterized in that the data collected by the transducer data are automatically transmitted to the stationary computer when a data connection from the buffer to the stationary computer. 15. The method according to any one of the preceding claims, characterized in that the signal emitted by the signal generator is emitted in the manner of a so-called bio-feedback signal. 16. Method according to one of the preceding claims, characterized in that the admissibility corridor is defined by movements carried out by the living being, and the data determining the permissibility corridor are stored in the evaluation electronics, wherein the same living being later uses the movement detection device, the evaluation electronics and the signal donors. 17. The method according to any one of the preceding claims, characterized in that in the evaluation electronics several admissibility Corridors are stored and different signals are triggered depending on which admissibility corridor was left by the automatically detected movements. 18. The method according to any one of the preceding claims, characterized in that in the evaluation electronics, a permissibility corridor is stored, which are typical, prior to a fall of the living beings motion patterns underlying. 19. A detection device for detecting movement data of living beings, wherein the movement detection device can be fastened to the living being or to an object moved by the living being, characterized by at least one movement-sensitive sensor which detects spatial movements, such as a 3-D Acceleration sensor, a 3-D magneto-sensitive sensor or a 3-D gyroscope, and a buffer which is associated with the detection device and receives the data detected by the transducer, and by an ID code identifying the detection device, which is stored in the detection device is, and by an interface, which enables a transmission of the sensor supplied by the movement data and the ID code to an external device, and wherein the detection device comprises an electronic evaluation circuit, which is configured evaluating the detected movement data according to the method of one of the preceding claims. 20. detection device according to claim 19, characterized in that the buffer is integrated into the detection device. 21. A detection device according to claim 19 or 20, characterized in that in addition to the 3-D sensor at least one further sensor is provided, which detects movement-relevant data, such as physiological data in the form of the pulse rate, respiratory rate or skin moisture of the animal, or as environmental parameters in the form of ambient temperature or ambient air pressure. 22 detection device according to one of claims 19 to 21, characterized in that it comprises at least two sensors, of which at least one is selectively switched on or off. 23. A detection device according to any one of claims 19 to 22, characterized by a clothing sensor, which is designed to cooperate with a provided on a garment signal generator, such that the clothing sensor generates a signal referred to as position signal, which defines the position at which the clothing sensor is arranged in the clothing of the athlete.