WO2025228791A1 - Radar sensor with self-learning for motion detection - Google Patents

Abstract

A method is provided for calibrating a radar motion sensor device located in a sensing space in which radar data signals are collected over a predetermined period of time. The radar data signals are processed to derive activity information, and that information is compared between portions of the predetermined time period. A type of the sensing space is thereby determined, wherein the type comprises an open plan space or a closed plan space. One or more thresholds for performing motion detection are set in dependence on the determined type of sensing space.

Description

RADAR SENSOR WITH SELF-LEARNING FOR MOTION DETECTION

FIELD OF THE INVENTION

The present invention generally relates to a method for calibrating a radar sensor for motion detection and a radar sensor comprising a control module configured to perform the method.

BACKGROUND OF THE INVENTION

Motion detectors are used in many applications ranging from burglar alarms to automatic door openers or automatic control of illumination or HVAC systems. For example, motion detectors are often used in offices, e.g. in conference rooms, to detect when people are present to steer light modules and/or HVAC systems such that lights, ventilation, heating and air conditioning are turned on when people are present and automatically turned off when no people are present. Automatic control of these systems based on motion detection has the potential to save large amounts of energy since e.g. the risk of lights or air conditioning units in an office remaining on during the night, or over the weekend, when no people are present is reduced.

Typically, Passive Infrared Sensors, PIRs, are used to detect motion. PIRs operate by registering incident IR radiation and when e.g., people move in front of the PIR the intensity distribution on the PIR sensor changes which can be taken as an indication that people are present in front of the sensor. A drawback with PIRs is, however, that they are not very sensitive and small motions, e.g. small hand gestures or small head motions, made by people sitting around a conference table may not always be detected by the PIR whereby the lights, or others systems, are deactivated while people are still using the conference room. In other words, the PIRs are prone to high false negative rates meaning that PIRs often fail to sense motion when there in fact is motion to detect.

In view of the drawbacks of PIRs, radar sensors have been used instead. Radar sensors are active sensors which transmit a radar signal into the environment and measure the reflected radar signal. If an object (e.g. a human) moves this will alter the reflected radar signal and this alteration can be used as an indication of motion. Radar sensors tend to be much more sensitive compared to PIRs such that even small motions can be detected which reduced the false negative rate. At the same time, radar signals can penetrate non-metal materials such as glass, plastic or drywall meaning that it is possible to hide radar motion sensors in walls or integrate the radar motion sensor into devices without the radar sensor being visible.

On the other hand, radar sensors are in some scenarios too sensitive leading to an increase in false positive rates. That is, radar sensors are prone to detecting motion even when no people are present. For example, since radar signals can penetrate drywall and glass, it is possible that motion in a completely separate room is detected by the radar sensor or that a person passing-by a room equipped with a radar sensor is registered by the radar sensor. To circumvent the issues with high false positive rates one common approach is to use a sensitivity setting that allows users to modify the sensitivity of the radar sensor. It has for example been proposed to automatically adapt sensitivity settings (in particular thresholds for motion detection).

Another issue is that different office applications require different motion sensing and control strategies. For example, in an open plan area, the motion sensor should be sensitive enough to cover a larger detection area, and thus fewer sensor devices would be required to realize the area control. For a meeting room, the sensor should be less sensitive to avoid false triggers from motions that are outside the room (e.g., passing-by along a corridor outside).

This conflict is again typically resolved during a pre-commissioning stage or by making use of manual settings to the radar sensor based on its installation locations, with different pre-set detection thresholds or strategies.

Some existing lighting control systems do not have an available method to perform the commissioning or manual settings to config, the radar sensor for a specific installation location. Thus, if installing a radar sensor that had been designed and optimized for an open plan installation to a meeting room, there would be many false triggers resulting from motion outside the meeting room. On the another hand, if installing a radar sensor that had been designed for a meeting room installation, to an open plan area, the sensor would be not sensitive enough and the detection range could be too small to be used.

As a result, two sensor designs need to be developed in order to cover both open plan areas and closed plan areas (e.g., meeting rooms), which increases the design and installation complexity. There is the potential for mis-installation with the wrong sensor type.

US2023184918A1 relates to a detection sensor including a signal transmitter configured to successively generate and transmit a detection signal in a plurality of observation intervals and a signal receiver configured to receive a scatter signal. The scatter signal includes a plurality of frequency bins resulting from a scattering of the detection signal in the plurality of observation intervals. The detection sensor includes an event generator configured to compare a present scatter signal to a previous scatter signal, separately for each of the plurality of frequency bins, and generate an event for each frequency bin, in which the present scatter signal differs from the previous scatter signal by at least an event threshold amount.

US2017123058A relates to detecting a presence of a person in an area of coverage using radar. A transmitter can transmit radio signals in a first direction in an area of coverage defined by a wall and a floor. A receiver can receive the transmitted radio signals reflected back from the area of coverage. A signal conditioning circuit can process the received radio signals. One or more hardware processors can be programmed to analyze the processed radio signals and detect a presence of a person in the area of coverage based on the analysis. The analysis of the processed signals can be performed in both time and frequency domain. In addition to radar, an input from an infrared sensor can also be used in conjunction with radar based detection.

US2024027599Alrelates to motion detection apparatuses. The motion detection may be performed using one or more of a sub-window of a predetermined time window, a predetermined threshold value that is settable responsive to changes in one or more environmental factors, or a detection trigger. An apparatus includes a processor and an analog-to-digital converter (ADC) circuitry to sample a reflected predetermined pattern signal to generate reflected predetermined pattern samples. The processor captures collections of the reflected predetermined pattern samples corresponding to a predetermined time window and determines a sum of the collections or sub-collections. The processor determines an average of magnitudes of the determined sum and determines that a moving object is detected responsive to a predetermined threshold value.

In view of these shortcomings there is a need for an improved method for calibrating a radar motion sensor and an improved radar motion sensor.

SUMMARY OF THE INVENTION

It is an object of the present invention to provide a new and improved method for calibrating a radar motion sensor. This and other objects are achieved by the method and radar motion sensor of the independent claims. Embodiments of the present invention are defined in the dependent claims. According to an aspect of the present invention, there is provided a method for calibrating a radar motion sensor device located in a sensing space, comprising: obtaining a temporal sequence of samples of a radar data signal, each sample indicating a first power level in a first frequency band of a radar signal detected by the radar motion sensor device; for a predetermined time period, collecting radar data signals; processing the radar data signals to derive activity information; comparing the derived activity information between portions of the predetermined time period and based on the comparison, determining a type of the sensing space; and setting one or more thresholds for performing motion detection in dependence on the determined type of sensing space.

The predetermined time period is preferably a plurality of days, and a portion of the time period is a day.

The invention thus provides a method that can determine the type of space, in particular office space, which is monitored by a radar sensor, by the radar sensor itself. Two types of office space are for example relevant: an open plan office space (such as multiple workstations in a shared room) and a closed plan office space (such as meeting rooms and private offices).

The type thus preferably comprises an open plan space or a closed plan space. An open plan space is a space with multiple workstations and optionally also communal open plan meeting areas. The space is shared between multiple workers so that multiple people move freely and frequently within the space. For example, the open plan space may include at least 5 workstations, or at least 10 workstations. A closed plan space is a space with a smaller number of fixed workstations, and the space is not shared with people other than those working at those workstations. A closed plan space for example has 4 workstations or less, or it may include one or two meeting tables. A meeting table may seat many people but they are all likely to move in and out of the space at the same time, and thus the movement pattern is similar to a single workstation.

After installation, the radar sensor starts a learning process to collect the characteristics of occupancy and activity in the monitored space over a predetermined time period, such as several working days. The underlying principle is that the characteristics of the two types of office space are different. In an open plan area, a radar sensor usually can cover multiple workstations, so the occupancy and activity are collectively determined by multiple users of these workstations. This means the variation of one or two people does not have a significant impact on the detected movement characteristics. For example, the monitored space is still occupied if 2 out of 5 people leave their workstations (e.g., for meetings). Therefore, the characteristics of occupancy and activity for open plan area are rather similar between different working days. On the other hand, the occupancy and activity levels in a closed plan space or meeting room could be rather different between different working days. For example, a meeting room may accommodate six 30 minutes meetings with two to three participants per meeting in one day, and three 1 hour meetings with four to six participants per meeting in another day.

Thus, in summary, the daily pattern of an open plan space is more dominated by the office hour rhythm, which is more or less similar day by day, whereas the daily pattern of a closed plan space is more dominated by the less predictable occupancy schedules day by day. For example, these may result from different meeting/occupancy sessions, occupancy times in one day, period of each occupancy, number of attendees during the occupancy. As a result, the similarity between daily patterns in these two types of office space are different.

Using the radar sensor data, a profile reflecting the activity in the space, such as the characteristics of occupancy and human activity in the monitored space, of a working day is created. After several profiles for several working days are generated, a comparison between the profiles is calculated. In one implementation, if a similarity is higher than a predefined threshold, the space is determined as an open plan area. Otherwise, the space is determined as a closed plan space. The learning is then completed and the motion detection method can be optimized for the determined space type.

The processing to derive activity information for example comprises: collecting signal power information over time; deriving a smoothed signal power by taking at least one moving average; selecting data sections of the smoothed signal power corresponding to active periods during the day; and normalizing the smoothed signal power for the active period data sections. This generates a characteristic waveform for each day, to enable a normalized comparison between days.

Deriving a smoothed signal power for example comprises deriving a difference between a first smoothed signal power smoothed over a first time period and a second smoothed signal power smoothed over a different second time period. In this way, a measure is provided that may take account of the rhythm of the overall daily occupancy as well as the level of occupancy change throughout the day.

The first time period is for example for generating a relatively short term smoothed signal and the second time period is for generating a relatively long term smoothed signal.

The first time period is for example in the range 1 to 3 hours and the second time period is in the range 6 to 12 hours. Thus, the smoothed signal power is a difference between a short-term smoothed signal (e.g. 2 hours) and a long-term smoothed signal (e.g. 8 hours).

Selecting data sections corresponding to active periods for example comprises setting a time window each side of the peak signal power for a smoothed signal power, smoothed over a time period in the range 6 to 12 hours.

This smoothed power signal may be the same long-term smoothed signal used to create a difference signal. However, it may be a different smoothed signal. It provides a peak at the center of the overall time window, and a fixed time window may be set to each side, such as 8 hours to each side, to make a full 16 hour day.

The step of normalizing for example comprises scaling the smoothed signal power such that the signal energy above a power threshold is the same for each day. This normalizing function makes the comparison between different days more reliable.

Comparing the derived activity information for example comprises obtaining a measure of similarity between the plurality of days. Various similarity measures may be used.

One example for obtaining a measure of similarity comprises generating a similarity matrix between activity information for pairs of days and deriving a similarity score from the similarity matrix.

If the measure of similarity is greater than a first threshold the sensing space may be determined to be an open plan space and if the measure of similarity is less than a second threshold the sensing space may be determined to be a closed plan space.

The first and second thresholds may be the same, so that spaces are categorized as open plan or closed plan. However, the first and second thresholds may different so that there is an additional region between the thresholds.

The invention also provides a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method defined above. The invention also provides a controller for a radar motion sensor, the controller comprising a processor and a memory, wherein the processor is configured to perform the method defined above.

BRIEF DESCRIPTION OF THE DRAWINGS

This and other aspects of the present invention will now be described in more detail, with reference to the appended drawings showing embodiments of the present invention.

Fig. 1 shows a top-down view of an exemplary room-layout with different types of office space;

Fig. 2 shows a method for calibrating a radar motion sensor device located in a sensing space;

Fig. 3 shows an example of the radar signals for a working week and shows the processing of step S3 of Fig. 2;

Fig. 4 is used to explain how the active period determination is carried out in more detail;

Fig. 5 is used to explain the normalization process;

Fig. 6 shows one example of data six days of activity information in a closed office space, in particular a meeting room; and

Fig. 7 shows another example set of waveforms from a radar sensor installed in the open area of an open office space.

DETAILED DESCRIPTION

The invention provides a method for calibrating a radar motion sensor device located in a sensing space in which radar data signals are collected over a predetermined time period, such as number of days. The radar data signals are processed to derive activity information, and that information is compared between portions of the time period, such as between the plurality of days. A type of the sensing space is thereby determined, wherein the type comprises an open plan space or a closed plan space. One or more thresholds for performing motion detection are set in dependence on the determined type of sensing space.

Fig. 1 shows a top down view of an exemplary indoor environment that introduces two types of office space. A first office space la is a closed plan space. In this type of space, one person or multiple people typically enter together and leave together. For example, the closed plan space may be a single office (with a single workstation 2) or it may be a meeting room with a meeting table. A second office space lb is an open plan space. It has multiple workstations 2. People typically come and go at different times, and there may be a communal area (e.g. drinks stations) where people come and go, even if they do not have a workstation in the space lb. It can be seen that the two office space types have very different activity profiles.

In the first space la there is a radar motion sensor 4a installed with the purpose of detecting motion in the first space la. For example, upon detection of motion in the first space la the lights are turned on in the first space la and/or the HVAC system associated with the first space la is activated. On the other hand, if no motion has been detected in the first space la for a predetermined period of time, the lights and/or HVAC system associated with the first room la is deactivated.

In the second space lb there is another radar motion sensor 4b, intended to detect motion at any of the workstations. There may of course be multiple sensors in a large open plan area. A corridor 1c passes the two office spaces.

The radar motion sensors 4a, 4b are sensitive and utilize electromagnetic signals capable of penetrating the partition walls or windows between the spaces. Thus, detection of motion in one area may take place for a radar sensor in another area.

To avoid these false positives, it is known to calibrate the radar motion sensors 4a, 4b in order to tune the sensitivity of the radar motion sensors such that the false positive rate can be reduced.

This tuning for example may take account of whether the sensing space is an open plan office space or a closed plan office space. This invention is based on the automatic detection of whether a space is open plan or closed plan, so that the calibration of the operation of the radar motion sensors can be performed automatically.

For example, for a close plan space, lighting needs to be turned on upon entry into the space and when there is movement in the space. A low sensitivity will prevent a false trigger caused by people passing outside the space, or people entering or staying in a neighboring closed plan space.

For an open plan area, the light can be triggered with far greater sensitivity.

For example, a walking motion within a designated distance, e.g., 3 meters, will trigger the lighting on (or to stay on), and a small motion, e.g. a keyboard activity or mobile phone activity within 2 meters range will still trigger the lighting on (or to stay on).

Fig. 2 shows a method for calibrating a radar motion sensor device located in a sensing space. The method comprises a step SI in which a temporal sequence of samples of a radar data signal are obtained. Each sample indicates a first power level in a first frequency band of a radar signal detected by the radar motion sensor device. The sensing may take place only in one frequency band but it may also take place in additional frequency bands.

Data is collected in step S2 for a plurality of days.

The data is then processed in step S3 to derive activity information for each day. The activity information captures information about the sensed movements in the sensing space, and this information is characteristic of the type of sensing space.

In step S4, the derived activity information is compared between the plurality of days, so that based on the comparison a type of the sensing space can be determined, wherein the type comprises an open plan space or a closed plan space.

In step S5, one or more thresholds for performing motion detection are set in dependence on the determined type of sensing space. Other control strategies may also be employed depending on the type of office space.

The steps will be explained below in greater detail.

Fig. 3 shows an example of the radar signal for a working week and shows the processing of step S3, applied to a closed plan working space, such as a meeting room.

The radar signal is provided to a processor which first performs analog to digital conversion using an ADC. A sampling frequency of 100Hz is for example used to obtain a digitized raw signal 10.

The time serial digital signal 10 is then processed via a fast Fourier transformation (FFT) and converted into a frequency domain signal for example in the range of [0Hz, 50Hz],

The low-frequency FFT result between 0Hz and 25Hz is then summed to calculate a single energy value labelled in the top graph as P(t). The average of the P(t) overall a sliding window of 15 minutes is calculated to obtain a single averaged energy value associated with a particular time point.

Thus, in step S3a, signal power information is determined over time.

The power signal P(t) result is further processed in step S3b which involves deriving a smoothed signal power by taking at least one moving average. In the example shown, a moving average of P(t) is calculated with a relatively short time period such as 2 hours (shown as PMA 2hr(t) in the top graph) and the moving average of P(t) is calculated with a relatively long time period such as 8 hours (shown as PMA 8hr(t) in the top graph).

The smoothed signal power in this example is finally generated as P’(t): P' (t) — PMA_2hr(t) ^— PMA_8hr (t)

Thus, the smoothed signal power comprises a difference between a first smoothed signal power smoothed over a first time period and a second smoothed signal power smoothed over a different second time period.

The calculated plot P’(t) is shown in the second graph in Fig. 3.

Thus, the signal P'(t) is first obtained by applying a moving average operation to the FFT data over a short period (15 minutes in this example) to derive a near- instantaneous average energy. Two longer moving average windows are then applied. The profile derived from the long moving average window (e.g., 8 hours) gives an index of a whole day occupancy rhythm, whereas the profile derived from the short moving average window (e.g., 2 hours) relates to occupancy sessions during the day. The difference between these two profiles characterizes the detailed occupancy schedules during the day.

As the Radar sensor is typically not able to obtain an absolute time from an external signal (e.g. via the internet), it is not possible to pre-determine the busy hours and the actual start/stop time for the daily profile extraction. By using a local timer, the inherent error will be accumulated and cause significant mismatch between the calculation and reality.

Thus, a method is also proposed to find the suitable time window from which the daily activity profile can be extracted that is well-matched with the busy hours and with no accumulated bias.

The method involves, in step S3c, selecting data sections of the smoothed signal power P'(t) that correspond to active periods during the day. These data sections P'(t)_i are shown in the third graph. There is also normalizing the smoothed signal power for the active period data sections in step S3d. This gives the Normalized P'(t)_i signals shown in the fourth graph.

The normalized data is then stored in a database 20 so that the comparisons (of step S4 in Fig. 1) can be made.

Fig. 4 is used to explain how the active period determination is carried out in more detail.

The top plot shows the profile of PMA _8hr(t), also shown in the top graph of Fig. 3. Again, this does not have to be the same plot as previously used, but could be a new plot.

From this averaged signal power waveform (smoothed over a time period in the range 6 to 12 hours) a daily peak point is found (shown by dots in the top graph of Fig. 4) and the corresponding time at which this peak is present. This time value will be considered to be a central time (tcenter) of the busy hours for each day.

A predefined busy hour time window is then set on each side of this daily center [t_center - 1 busy hour, t_center + busy hour].

For the example shown in Fig. 4, the busy hour time window is set as 16 hours but it can be changed to any other desired setting. The 16 hour daily busy hour time window can then be used to section the daily portion of the signal P’(t) and become the signal P’(t)_i shown in the third graph of Fig. 3.

All the daily data that has been sectioned to form the P’ (t)_i signals is then further processed to provide the normalization shown in the fourth graph of Fig. 3.

Fig. 5 is used to explain the normalization process.

The top plot again shows the profile of PMA 8hr(t), as already shown in the top graph of Fig. 3. From this averaged signal power waveform (smoothed over a time period in the range 6 to 12 hours), a threshold is defined, which is higher than the average value of PMA 8hr(t) outside the defined busy hour time windows. The threshold is shown as T P in the top graph of Fig. 5. The area under the graph, above this threshold, is calculated for each daily portion of PMA 8hr(t) and defined as the Weight i (i here stands for the 1^th day data used during the self-learning).

P’(t)_i, shown in the second graph of Fig. 5, is then normalized as:

P'(t)_i

Normalized P'(t) i = — — : — Weight.!

This means the smoothed signal power P'(t)_i is normalized such that the signal energy above a power threshold is the same for each day.

A numerical and normalized profile of Normalized P’ (t)_i is thereby obtained as shown in the bottom graph of Fig. 5. Each daily profile of the plot Normalized P’(t)_i reflects the characteristics of occupancy and human activity in the monitored space of that specific day. These characteristics are referred to as "activity information" in this disclosure.

Each daily Normalized P’(t)_i signal comprises a ID array of processed results with a fixed data length that is determined by the busy hour time window length. All the ID array values of Normalized P’(t)_i that are processed and calculated during the self-learning process are stored in the internal database 20 of the radar sensor. Once the sensor has obtained sufficient daily Normalized P’(t)_i profiles, the day-to-day similarity of these normalized profiles is calculated, as shown in step S4 of Fig. 2.

There are many known methods that can be used to perform data-array based similarity calculations. One option is to generate a similarity matrix via a paired comparison between different days.

Fig. 6 shows one example of data for six days of activity information in a closed office space, in particular a meeting room. It corresponds to the waveforms of Fig. 2. The diagonal has values 1 (because it is a comparison of identical days). The other values represent comparisons of pairs of days. For example, the middle row, illustrated in a box, shows a comparison of day 4 with all of days 1 to 6.

The overall similarity will be in the range of [0,1] and can be calculated as: yi-6,j- ji=i = i lScorejj l

Similarity =

36

Thus, it is an average similarity for all comparisons. In this example, the overall similarity is 0.45.

Fig. 7 shows another example set of waveforms from a radar sensor installed in the open area of an open office space. It shows the same plots as in Fig. 2. The similarity calculated in this case higher, and can reach 0.96.

Thus, if the measure of similarity is greater than a first threshold the sensing space can be determined to be an open plan space and if the measure of similarity is less than a second threshold the sensing space is determined to be a closed plan space.

By way of example, a similarity of less than 0.6 can be used to conclude a closed plan office space while a similarity of greater than 0.8 can be used to conclude an open plan office space. Of course, other threshold settings can be used.

The method has been tested by installing six radar sensors in different office environments. As demonstrated by the similarity values shown in the table below, by using the method above, the sensors themselves can self-learn their installation location very well.

As explained above, different radar sensor sensitivities or even different sensing strategies may be applied to the different types of sensing space.

In respect of the closed plan space (e.g. meeting room), a threshold power PO may be associated with a person passing along a corridor outside an office space, whereas a higher power Pl is associated with entry into the office space and movement within the office space. The threshold for turning on the light may be P_TH = P0+s(Pl-P0) where s is a sensitivity factor. The sensitivity factor for example has a default value (e.g., 0.8) but it can be adjusted.

In respect of an open plan space, a mechanism may be employed to ensure more sensitive motion detection. For example, a statistical parameter or parameters may be derived from the radar detection power, such as the mean and/or standard deviation over a previous period, such as three minutes. The statistical parameter or parameters may be compared with corresponding historical data, to determine whether to update the thresholds for triggering on the lights, holding the lights on, and turning off the lights.

Thus, the determination of whether a sensing space is an open plan space or a closed plan space can be used not simply to adjust threshold levels, but also to set a control strategy for the office space.

The applicant has proposed, but not yet published, a method for determining the detected power levels associated with movement in an office space compared to movement outside the office space (e.g. a passing-by event along a corridor). The method is also used as part of a learning phase for setting the thresholds that will be used to identify movement in a space using a radar sensor. The one or more thresholds are selected such that passing-by events are not falsely recognized as movement in the room being monitored. This is achieved by detecting passing-by events during the learning phase by means of a pattern recognition algorithm. During the learning phase, this pattern recognition algorithm can take account of the radar data signal samples before and after a detection event, and this enables passing-by events to be detected.

A passing by event may involve a person walking past an entrance to the room, e.g. along a corridor, or it may be a person walking along a glass wall or stud plasterboard wall of the room.

To detect such a passing by event, it is appreciated that a person walking along a corridor past the space being monitored will approach the sensor (the distance between the radar and the person will keep decreasing) until the person is positioned nearest the sensor (but still outside the space being monitored). The distance then starts increasing as the person walks away from the space being monitored. Accordingly, the power of the radar signal reflecting the motion of the walking person firstly increases to a maximum then decreases, and the increasing and the decreasing are roughly symmetrical. It is proposed to use a pattern recognition algorithm to detect such a motion pattern and thereby learn the suitable trigger thresholds for such a passing-by event (described as P0 above) and for movement into and in the space being monitored (described as Pl above). The threshold to be applied can then be set between these values using the sensitivity parameter described above, for example for a closed office lighting control space.

The radar sensor used by the method above may use any type of Doppler radar, such as a pulsed Doppler radar or an unmodulated continuous wave (CW) Doppler radar. In the latter example, a constant frequency carrier wave in the GHz range is emitted and its reflection received. Transmitted and reflected waves are mixed to form an intermediate frequency (IF) signal in the time domain which is outputted as the “raw” radar signal. In available radar motion sensors, the carrier frequency may be for example 5.8 GHz, 24 GHz or 60 GHz. Accordingly, the radar signal comprises a plurality of successive samples, with each sample representing the radar signal at a respective point in time.

The frequency content of a radar signal may be extracted in one or more frequency bands forming a radar data signal having a temporal sequence of samples with each sample representing the spectral power in the one or more frequency bands. For example a low frequency band may contain frequencies 0 - 4 Hz and a high frequency band may contain frequencies 6 - 10 Hz. However, it is understood that these frequency bands are merely exemplary and while these frequency bands are suitable for radar motion sensors using 5.8 GHz carrier signals, other frequency bands may be used if other carrier frequencies are used. When multiple frequencies are used, each sample in the radar data signal may be represented with a vector associated with an individual timestamp, wherein each vector element represents the spectral power in a respective frequency band.

The person skilled in the art realizes that the present invention by no means is limited to the preferred embodiments described above. On the contrary, many modifications and variations are possible within the scope of the appended claims. For example, it is possible to use one the spectral power in one frequency band or two frequency bands. Additionally, it is envisaged that the spectral power in more than two frequency bands may be used to perform the calibration.

Claims

CLAIMS:

1. A method for calibrating a radar motion sensor device located in a sensing space, comprising:

(51) obtaining a temporal sequence of samples of a radar data signal, each sample indicating a first power level in a first frequency band of a radar signal detected by the radar motion sensor device;

(52) for a predetermined time period, collecting radar data signals;

(53) processing the radar data signals to derive activity information; and

(54) comparing the derived activity information between portions of the predetermined time period and based on the comparison, determining a type of the sensing space; wherein comparing the derived activity information comprises obtaining a measure of similarity between the portions of the predetermined time period; and if the measure of similarity is greater than a first threshold, the sensing space is determined to be an open plan office space and if the measure of similarity is less than a second threshold, the sensing space is determined to be a closed plan office space.

2. The method of claim 1, wherein the type of the sensing space comprises an open plan space or a closed plan space.

3. The method of claim 1 or 2, wherein the time period comprises a plurality of days, and each portion of the time period is a day.

4. The method of any one of claims 1 to 3, wherein the processing to derive activity information comprises:

(S3 a) determining signal power information over time;

(S3b) deriving a smoothed signal power by taking at least one moving average;

(S3c) selecting data sections of the smoothed signal power corresponding to active periods during the portion of the predetermined time period; and (S3 d) normalizing the smoothed signal power for the active period data sections.

5. The method of claim 4, wherein (S3b) deriving a smoothed signal power comprises deriving a difference between a first smoothed signal power smoothed over a first time period and a second smoothed signal power smoothed over a different second time period.

6. The method of claim 5, wherein the first time period is for generating a relatively short term smoothed signal and the second time period is for generating a relatively long term smoothed signal.

7. The method of claim 6, wherein the first time period is the in the range 1 to 3 hours and the second time period is in the range 6 to 12 hours.

8. The method of any one of claims 4 to 7, wherein the (S3c) selecting data sections corresponding to active periods comprises setting a time window each side of the peak signal power for a smoothed signal power, smoothed over a time period in the range 6 to 12 hours.

9. The method of any one of claims 4 to 8, wherein the (S3d) normalizing comprises scaling the smoothed signal power such that the signal energy above a power threshold is the same for each day.

10. The method of any one of claims 1 to 9, wherein obtaining a measure of similarity comprises generating a similarity matrix between activity information for pairs of the portions of the predetermined time period and deriving a similarity score from the similarity matrix.

11. The method of any one of claims 1 to 10, wherein the method further comprises:

(S5) setting one or more thresholds for performing motion detection in dependence on the determined type of sensing space.

12. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method according to any of the preceding claims. 13. A controller for a radar motion sensor, the controller comprising a processor and a memory, wherein the processor is configured to perform the method according to any of claims 1 to 11.