CN117409538A - Wireless anti-fall alarm system and method for nursing care - Google Patents

Abstract

The invention discloses a wireless anti-fall alarm system and method for nursing care, and relates to the field of intelligent alarm technology. It uses a camera to collect behavioral videos of the elderly and transmits the videos to a background anti-fall server through a wireless transmission module. More specifically, in the background anti-fall server, artificial intelligence technology based on deep learning is used to process and analyze behavioral monitoring videos to identify the behavioral characteristics of the elderly and intelligently determine whether to generate anti-fall alarm prompts. In this way, when it is detected that the elderly person's behavior may pose a risk of falling, an early warning prompt will be generated through the calling device in time to notify relevant personnel for rescue or intervention.

Description

Wireless anti-fall alarm system and method for nursing care

Technical field

The present invention relates to the field of intelligent alarm technology, and in particular to a wireless anti-fall alarm system and method for nursing care.

Background technique

With the trend of population aging, the health and safety issues of the elderly have received increasing attention from society. Due to the degradation of physical functions, the elderly are prone to accidents such as falls, which may bring a huge burden to themselves and their families.

Currently, there are some products on the market for fall prevention, such as fall detectors. These products usually require the elderly to wear or carry specific equipment, such as bracelets, belts, watches, etc. However, these devices have some shortcomings, such as uncomfortable wearing, short battery life, and high false alarm rate.

Therefore, an optimized solution is expected.

Contents of the invention

Embodiments of the present invention provide a wireless anti-fall alarm system and method for nursing care, which uses a camera to collect behavioral videos of the elderly and transmits the videos to a background anti-fall server through a wireless transmission module. More specifically, in the background anti-fall server, artificial intelligence technology based on deep learning is used to process and analyze behavioral monitoring videos to identify the behavioral characteristics of the elderly and intelligently determine whether to generate anti-fall alarm prompts. In this way, when it is detected that the elderly person's behavior may pose a risk of falling, an early warning prompt will be generated through the calling device in time to notify relevant personnel for rescue or intervention.

An embodiment of the present invention also provides a wireless anti-fall alarm method for nursing, which includes:

Obtain behavioral monitoring videos of elderly subjects being monitored collected by cameras;

Transmit the behavior monitoring video to the backend anti-fall server through the wireless transmission module;

In the background anti-fall server, the behavior monitoring video is analyzed and processed to determine whether an anti-fall alarm prompt is generated;

In response to generating an anti-fall alarm prompt, an early warning prompt is generated through the calling device;

Among them, in the background anti-fall server, the behavior monitoring video is analyzed and processed to determine whether an anti-fall alarm prompt is generated, including:

Perform data preprocessing on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames;

Extract the timing characteristics of the elderly behavior in the sequence of clipped behavioral key frames to obtain the timing change feature vector of the behavior of the elderly;

Based on the time-series change feature vector of the elderly person's behavior, determine whether to generate an anti-fall alarm prompt;

Wherein, data preprocessing is performed on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames, including:

Perform discrete sampling processing on the behavior monitoring video to obtain a sequence of behavior monitoring key frames;

The sequence of behavioral monitoring key frames is randomly trimmed to obtain the sequence of trimmed behavioral key frames.

An embodiment of the present invention also provides a wireless anti-fall alarm system for nursing care, which includes:

Surveillance video collection module, used to obtain behavioral surveillance videos of elderly subjects being monitored collected by cameras;

A video transmission module, used to transmit the behavior monitoring video to the background anti-fall server through the wireless transmission module;

An analysis and processing module, used in the background anti-fall server to analyze and process the behavior monitoring video to determine whether an anti-fall alarm prompt is generated;

An early warning prompt generation module is used to generate an early warning prompt through the calling device in response to generating an anti-fall alarm prompt;

Among them, the early warning prompt generation module includes:

Perform data preprocessing on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames;

Extract the timing characteristics of the elderly behavior in the sequence of clipped behavioral key frames to obtain the timing change feature vector of the behavior of the elderly;

Based on the time-series change feature vector of the elderly person's behavior, determine whether to generate an anti-fall alarm prompt;

Wherein, data preprocessing is performed on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames, including:

Perform discrete sampling processing on the behavior monitoring video to obtain a sequence of behavior monitoring key frames;

The sequence of behavioral monitoring key frames is randomly trimmed to obtain the sequence of trimmed behavioral key frames.

Beneficial effects of the present invention:

Use cameras to collect behavioral videos of the elderly, and transmit the videos to the backend anti-fall server through the wireless transmission module. In the backend anti-fall server, artificial intelligence technology based on deep learning is used to process and analyze the behavioral monitoring videos to identify the elderly. Behavioral characteristics, and intelligently determine whether to generate an anti-fall alarm prompt. In this way, when it is detected that the elderly person's behavior may pose a risk of falling, an early warning prompt is generated through the calling device in a timely manner to notify relevant personnel for rescue or intervention.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without exerting creative efforts. In the attached picture:

Figure 1 is a flow chart of a wireless anti-fall alarm method for nursing provided in an embodiment of the present invention.

FIG. 2 is a schematic diagram of the system architecture of a wireless anti-fall alarm method for nursing provided in an embodiment of the present invention.

Figure 3 is a block diagram of a wireless anti-fall alarm system for nursing provided in an embodiment of the present invention.

Figure 4 is an application scenario diagram of a wireless anti-fall alarm method for nursing provided in an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not used to limit the present invention.

Unless otherwise stated, all technical and scientific terms used in the embodiments of this application have the same meanings as commonly understood by those skilled in the technical field of this application. The terminology used in this application is for the purpose of describing specific embodiments only and is not intended to limit the scope of this application.

In the description of the embodiments of this application, it should be noted that, unless otherwise stated and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection, or it can be an internal connection between two elements, or it can be a direct connection. , or can be indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

It should be noted that the terms "first\second\third" involved in the embodiments of this application are only used to distinguish similar objects and do not represent a specific ordering of objects. It is understandable that "first\second\third" Three" may interchange specific order or precedence where permitted. It is to be understood that the "first\second\third" distinction may be interchanged under appropriate circumstances so that the embodiments of the present application described herein may be practiced in sequences other than those illustrated or described herein.

With the trend of population aging, the health and safety issues of the elderly have received increasing attention from society. Due to the degradation of physical functions, the elderly are prone to accidents such as falls, which may bring a huge burden to themselves and their families. First of all, the health and safety issues of the elderly are of great significance to both individuals and families. The deterioration of the body functions of the elderly makes them more susceptible to injury. In particular, falls have become a common accident. Falls may lead to serious consequences such as fractures and internal bleeding. Even life-threatening. In addition, the recovery process of the elderly is also relatively slow and may require long-term care and rehabilitation treatment. This not only puts a physical and psychological burden on the elderly themselves, but also puts economic and time pressure on their families.

In order to solve the health and safety problems of the elderly, many products have appeared on the market that require the elderly to wear or carry specific equipment, such as bracelets, belts, watches, etc. However, there are some disadvantages to these devices, such as: some older people may find these devices uncomfortable to wear, especially for those with arthritis or sensitive skin, and inappropriate wearing methods may cause wear, scratches, or allergies. problems, thereby reducing the willingness and ability of older adults to use these devices. Many devices require regular charging, and older adults may forget to do so due to memory loss or difficulty operating them. In addition, the battery life of some devices is relatively short and requires frequent charging, which also brings inconvenience to the elderly. Due to technical limitations or sensor inaccuracies, some devices may have false alarm issues, meaning the device may mistakenly trigger alarms or notifications, causing unnecessary panic and distress to seniors and family members.

To address these issues, improvements in technology and design can improve the experience and effectiveness of these devices for the elderly. For example, the wearing method of the devices can be optimized and soft, breathable materials can be used to ensure comfort and skin-friendliness. At the same time, extend battery life and optimize charging methods, such as using portable chargers or wireless charging technology, to reduce the charging burden on the elderly. In addition, sensor technology and algorithms are continuously improved to reduce false alarm rates and improve equipment accuracy and reliability.

In addition to improving the devices themselves, there are other ways to provide health and safety monitoring for older adults. For example, smart home technology can be used to install sensors and cameras to monitor the activities and behaviors of the elderly, as well as provide remote monitoring and emergency rescue. This solution that does not require the wearing of specific equipment may be more convenient and acceptable, and can also reduce the risk of elderly people. Human discomfort and privacy concerns.

In one embodiment of the present invention, FIG. 1 is a flow chart of a wireless anti-fall alarm method for nursing provided in the embodiment of the present invention. FIG. 2 is a schematic diagram of the system architecture of a wireless anti-fall alarm method for nursing provided in an embodiment of the present invention. As shown in Figures 1 and 2, the wireless anti-fall alarm method for nursing care according to the embodiment of the present invention includes: 110, obtaining the behavior monitoring video of the monitored elderly subject collected by the camera; 120, converting the behavior monitoring video Transmitted to the background anti-fall server through the wireless transmission module; 130, in the background anti-fall server, analyze and process the behavior monitoring video to determine whether an anti-fall alarm prompt is generated; and, 140, in response to the generation of an anti-fall alarm Prompt, generate early warning prompt through calling device.

In step 110, the behavior monitoring video of the monitored elderly subject collected by the camera is obtained. When setting up the camera, ensure that its position and angle can fully cover the activity area of the monitored elderly person to obtain accurate behavioral monitoring video. At the same time, we must respect the privacy rights of the elderly and ensure that the camera settings do not infringe on their personal privacy. By obtaining behavioral monitoring videos, the activities and behaviors of the elderly can be observed in real time and potentially dangerous situations, such as falls and loss of balance, can be detected in a timely manner.

In step 120, the behavior monitoring video is transmitted to the background anti-fall server through the wireless transmission module. When performing wireless transmission, ensure the stability and security of the transmission to avoid loss of video data or access by unauthorized persons. At the same time, appropriate wireless transmission technology and equipment must be selected to ensure transmission effect and speed. By wirelessly transmitting behavioral monitoring videos to the backend anti-fall server, real-time monitoring and data storage can be achieved to facilitate subsequent analysis and processing.

In step 130, the background anti-fall server analyzes and processes the behavior monitoring video to determine whether an anti-fall alarm prompt is generated. When analyzing and processing behavioral monitoring videos, advanced image processing and computer vision algorithms are used to accurately identify the behavior and movements of the elderly and determine whether there is a risk of falling. At the same time, it is necessary to ensure the accuracy and stability of the system and minimize the probability of false alarms. By analyzing and processing behavioral monitoring videos, the behavioral characteristics of the elderly, such as falling, falling, sudden cessation of activities, etc., can be monitored in real time, and early warnings can be issued in a timely manner to improve the accuracy and effectiveness of fall prevention.

In step 140, in response to generating an anti-fall alarm prompt, a pre-warning prompt is generated by calling the device. When responding to the anti-fall alarm prompt, ensure that the setting and operation of the calling device are simple and easy to use, so that the elderly can send a signal for help in a timely manner. At the same time, an effective alarm response mechanism must be established to ensure that relevant personnel can receive and respond to early warning prompts in a timely manner. The early warning prompt generated by the calling device can quickly notify relevant personnel, such as family members, caregivers or medical staff, so that they can take timely action to provide emergency assistance and care, and reduce the occurrence and serious consequences of fall accidents.

The above steps play an important role in monitoring the health and safety of the elderly. During the implementation process, pay attention to the setting and operation of the equipment, ensure the security and privacy of the data, and establish an effective alarm response mechanism. In this way, the behavior of the elderly can be monitored in real time, the risk of falls can be accurately determined, and early warning prompts can be issued in a timely manner, thereby improving the safety and health of the elderly.

In response to the above technical problems, the technical concept of this application is to use a camera to collect behavioral videos of the elderly and transmit the videos to the background anti-fall server through a wireless transmission module. More specifically, in the background anti-fall server, artificial intelligence technology based on deep learning is used to process and analyze behavioral monitoring videos to identify the behavioral characteristics of the elderly and intelligently determine whether to generate anti-fall alarm prompts. In this way, when it is detected that the elderly person's behavior may pose a risk of falling, an early warning prompt will be generated through the calling device in time to notify relevant personnel for rescue or intervention.

Based on this, in the technical solution of this application, the behavior monitoring video of the monitored elderly subject collected by the camera is first obtained; and the behavior monitoring video is transmitted to the background anti-fall server through the wireless transmission module. That is, using video data to understand the behavioral patterns and dynamic changes of the elderly. It should be understood that videos can provide intuitive, detailed, and continuous information, including the elderly's posture, movements, gait, etc., which can help analyze and judge the behavioral status of the elderly. Among them, transmitting the collected behavioral monitoring videos to the background anti-fall server can centrally manage and store a large amount of monitoring data, avoiding the problem of large-scale data storage on the device side.

Next, in the background anti-fall server, the behavior monitoring video is discretized and sampled to obtain a sequence of behavior monitoring key frames; and the sequence of behavior monitoring key frames is randomly trimmed to obtain the trimmed behavior key frames. the sequence of. Here, behavioral monitoring videos are usually continuous video streams, which contain a large amount of redundant information. In order to reduce the cost of calculation and storage, and reduce the complexity of subsequent processing, in the technical solution of this application, discrete sampling processing is performed to extract key information in the video. In addition, random clipping of individual behavioral monitoring keyframes can reduce the effects of noise, occlusion, and other environmental changes.

In a specific embodiment of the present application, in the background anti-fall server, the behavior monitoring video is analyzed and processed to determine whether an anti-fall alarm prompt is generated, including: performing data preprocessing on the behavior monitoring video to Obtain a sequence of clipped behavioral key frames; extract the timing characteristics of the elderly behavior in the sequence of clipped behavioral key frames to obtain a timing change feature vector of the behavior of the elderly; and, based on the timing variation feature vector of the behavior of the elderly, determine whether an anti-prevention event occurs. Fall alarm prompt.

Among them, through data preprocessing and clipping, key frame sequences can be extracted from behavioral monitoring videos, that is, image frames containing important actions and behaviors. This can reduce the amount of data, improve the efficiency of subsequent analysis, and focus on key behaviors, thereby Determine behavioral changes of the elderly more accurately.

Then, by extracting the timing features of the elderly behavior in the clipped behavioral key frame sequence, the evolution and change patterns of the elderly behavior can be captured. These features can include information such as gait, posture, activity frequency, and movement amplitude. By combining these features into feature vectors, the behavioral status and changing trends of the elderly can be better described, providing useful information for subsequent analysis and decision-making.

Then, through analysis based on the time-series change feature vector of the elderly's behavior, it can be determined whether the elderly is at risk of falling. For example, models or machine learning algorithms can be built to identify abnormal behavioral patterns, such as sudden pauses, irregular gaits, etc., as well as characteristics associated with falls. When these abnormal behaviors are detected, the system can generate anti-fall alarm prompts and promptly notify relevant personnel to take necessary rescue measures, thereby reducing the incidence of fall accidents.

By preprocessing and tailoring behavioral monitoring videos, extracting the timing characteristics of the elderly's behavior, and judging whether to generate anti-fall alarm prompts based on these characteristics, the accuracy and effectiveness of the elderly health and safety monitoring system can be improved.

Further, in a specific embodiment of the present application, performing data preprocessing on the behavior monitoring video to obtain a sequence of trimmed behavior key frames includes: performing discretization sampling processing on the behavior monitoring video to obtain the behavior monitoring A sequence of key frames; and, randomly trimming the sequence of behavioral monitoring key frames to obtain the sequence of trimmed behavioral key frames.

Among them, discrete sampling processing can convert continuous behavioral monitoring videos into a series of discrete key frames. By selecting key frames for processing and analysis, the amount of data and computing load can be reduced, and the efficiency of the system can be improved. Key frames usually contain important action and behavior information, so the key behaviors of the elderly can be more focused and analyzed to more accurately determine whether to generate an anti-fall alarm prompt.

Random clipping can select different time periods and perspectives from the sequence of behavioral monitoring key frames to obtain diverse behavioral key frames, which can increase the diversity of data and improve the generalization ability and robustness of the model. Through the tailored behavioral key frame sequence, different aspects and changing patterns of the elderly's behavior can be captured more comprehensively, thereby more accurately extracting behavioral characteristics and determining whether to generate anti-fall alarm prompts.

Perform discrete sampling processing on behavior monitoring videos to obtain a sequence of behavior monitoring key frames, and randomly trim the key frame sequence to obtain a sequence of trimmed behavior key frames, which can improve system efficiency, reduce computing load, and increase data efficiency. Diversity and generalization capabilities can more accurately capture the key information and changing patterns of the elderly's behavior, providing a more useful data basis for subsequent feature extraction and classification analysis.

In a specific example of this application, the sequence of behavior monitoring key frames is randomly clipped based on the area of interest of the elderly. That is, if the area where the elderly is located accounts for more than 80% of the clipping area, it is retained. In this way, the influence of the background can be removed to a certain extent, allowing the model to focus on learning the behavioral patterns of the elderly area.

Then, the sequence of clipped behavioral key frames is passed through a behavioral feature extractor based on a convolutional neural network model to obtain a sequence of behavioral feature vectors of the elderly. That is, a convolutional neural network model is used to construct the behavioral feature extractor to capture the behavioral patterns and temporal dynamic changes of the elderly.

In a specific embodiment of the present application, extracting the elderly behavior temporal characteristics in the sequence of clipped behavioral key frames to obtain the elderly behavior temporal change feature vector includes: using a deep learning network model to extract the clipped behavioral key frames The behavioral characteristics of the elderly in the sequence to obtain the sequence of behavioral characteristic vectors of the elderly; and, calculate the correlation between the first N-1 elderly behavioral characteristic vectors and the Nth elderly behavioral characteristic vector in the sequence of the elderly behavioral characteristic vectors. degree to obtain the time-series change feature vector of the elderly behavior composed of multiple correlation degrees. N is the number of the elderly behavior feature vectors in the sequence of the elderly behavior feature vectors.

A deep learning network model is used to extract the behavioral features of the elderly in the sequence of clipped behavioral key frames to obtain a sequence of behavioral feature vectors of the elderly. Among them, the deep learning network model can be used to learn and extract rich feature representations in the clipped behavioral key frames. The deep learning model can automatically learn and capture the semantic information and contextual relationships in the image, thereby more accurately characterizing the behavior of the elderly. By extracting the sequence of the elderly's behavior feature vectors, the behavior can be converted into a computer-processable form, providing a basis for subsequent time series analysis.

Calculate the correlation between the first N-1 elderly behavior feature vectors and the Nth elderly behavior feature vector in the sequence of the elderly's behavior feature vectors to obtain the time-series change feature vector of the elderly's behavior composed of multiple correlations. . By calculating the correlation between adjacent vectors in the behavioral feature vector sequence, the temporal change pattern of the elderly behavior can be captured. The correlation reflects the similarity and changing trend between adjacent behavioral feature vectors, and can be used to judge the stability and abnormality of the elderly's behavior. By combining multiple correlation degrees into a time-series changing feature vector, the dynamic changes in the elderly's behavior can be described more comprehensively, and more information can be provided to determine whether to generate an anti-fall alarm prompt.

Using the deep learning network model to extract the sequence of the elderly's behavioral feature vectors and calculating the correlation to obtain the temporal change feature vectors of the elderly's behavior can improve the sensitivity and accuracy of the changes in the elderly's behavior. In this way, rich behavioral feature representations can be extracted, the temporal change patterns of behavior can be captured, and more information can be provided to judge the stability and abnormality of the elderly's behavior to further improve the performance and effectiveness of the elderly health and safety monitoring system.

Wherein, the deep learning network model is a behavioral feature extractor based on a convolutional neural network model; wherein the behavioral feature extractor based on a convolutional neural network model includes: an input layer, a convolution layer, a pooling layer, Activation layer and output layer.

Specifically, using a deep learning network model to extract the behavioral features of the elderly in the sequence of clipped behavioral key frames to obtain a sequence of behavioral feature vectors of the elderly includes: passing the sequence of clipped behavioral key frames through the convolution-based The behavioral feature extractor of the neural network model is used to obtain the sequence of behavioral feature vectors of the elderly.

The convolutional neural network is a powerful deep learning model that is good at extracting high-level features from image data. By inputting the clipped behavioral key frame sequence into the convolutional neural network model, it can automatically learn and extract abstract features related to the behavior of the elderly. , these features can capture the spatial and temporal structure of behavior, including information on posture, action patterns, speed, etc., thereby more comprehensively describing the behavioral characteristics of the elderly.

By inputting the behavioral key frame sequence into the convolutional neural network model, the temporal change pattern of the behavior can be captured. The convolutional neural network model can use operations such as convolution and pooling to extract timing-related features when processing sequence data. This can better understand the dynamic changes in the elderly's behavior, including the duration of behavior, speed changes, the order of behavior sequences, etc. Time series change modeling helps to more accurately determine whether to generate anti-fall alarm prompts, and to Continuous monitoring and analysis of the elderly’s behavioral status.

By using a behavioral feature extractor based on a convolutional neural network model, the feature extraction process can be automated. Traditional feature extraction methods require manual definition and extraction of features, while feature extractors based on convolutional neural networks can automatically learn and extract task-related features, which can reduce the workload and subjectivity of manual feature engineering and improve the accuracy of features. sexuality and expressiveness.

Subsequently, the correlation degree between the first N-1 elderly behavior feature vectors and the Nth elderly behavior feature vector in the sequence of the elderly behavior feature vectors is calculated respectively to obtain the elderly behavior time series change feature vector composed of multiple correlation degrees. , where N is the number of the elderly behavior feature vectors in the sequence of the elderly behavior feature vectors. It should be understood that the behavior of the elderly usually has certain temporal change characteristics, and by calculating the correlation between adjacent feature vectors, the behavior change trend over time can be captured. This helps to understand the daily activity patterns, behavioral habits and possible abnormal behaviors of the elderly. Specifically, a low correlation value may mean that an older person has experienced unusual changes in behavior, such as a fall, coma, or other emergency.

Further, the time-series change feature vector of the elderly person's behavior is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether an anti-fall alarm prompt is generated; and in response to the classification result indicating that an anti-fall alarm prompt is generated, by calling The device generates an early warning prompt. Among them, the calling device is usually used to send emergency requests or calling signals to others to obtain help or support. Calling devices are commonly found in medical institutions, nursing homes, security centers, home environments and other places where timely response to emergencies is required. They can provide a convenient way for the elderly, sick, disabled, etc. to seek help and provide timely assistance in emergencies. Here, after the calling device generates an early warning prompt, the relevant nursing staff can take corresponding measures in time to prevent the elderly from accidents.

In a specific embodiment of the present application, determining whether to generate an anti-fall alarm prompt based on the time-series change feature vector of the elderly person's behavior includes: passing the time-series change feature vector of the old man's behavior through a classifier to obtain a classification result, and the classification The result is used to indicate whether an anti-fall alarm prompt is generated.

By inputting the time-series change feature vector of the elderly's behavior into the classifier for classification, the behavioral status of the elderly can be monitored in real time. The classifier can determine whether the current behavioral feature vector belongs to a fall or other dangerous behavior based on the pre-trained model and labeled data. category, if the classification results indicate that a fall or other dangerous situation may occur, the system can immediately generate an anti-fall alarm prompt so that necessary rescue measures can be taken in a timely manner. By using a classifier to classify the time-series change feature vectors of the elderly's behavior, an automated decision-making process can be realized. The classifier can automatically determine the behavior status of the elderly based on patterns and rules, and trigger corresponding alarm prompts as needed, which can alleviate manual monitoring. burden and improve the efficiency and accuracy of the monitoring system.

The classifier can be personalized according to the behavioral characteristics and needs of different elderly people. By using diversified data and features through the classifier, the classifier can better adapt to the behavioral patterns and habits of different elderly people, which can improve the flexibility of the system. and adaptability, more accurately determine whether to generate an anti-fall alarm prompt, and reduce false alarms and false negatives. Classify the time-series change feature vector of the elderly's behavior through a classifier to obtain the classification result to indicate whether an anti-fall alarm prompt is generated. This can achieve the beneficial effects of real-time warning, automated decision-making and personalized adaptation to improve the health and safety monitoring system for the elderly. accuracy, efficiency and user experience.

In one embodiment of the present application, the wireless anti-fall alarm method for nursing also includes a training step: training the behavioral feature extractor and the classifier based on the convolutional neural network model; wherein, The training step includes: obtaining training data, the training data including training behavior monitoring video, and whether the true value of the anti-fall alarm prompt is generated; in the background anti-fall server, discretely performing the training behavior monitoring video Sampling processing is performed to obtain a sequence of training behavior monitoring key frames; the sequence of training behavior monitoring key frames is randomly trimmed to obtain a sequence of training behavior monitoring key frames; and the sequence of training behavior monitoring key frames is passed through the Describe the behavioral feature extractor based on the convolutional neural network model to obtain the sequence of training old man behavior feature vectors; calculate the first N-1 training old man behavior feature vectors and the Nth training old man behavior feature vector in the sequence of training old man behavior feature vectors. The correlation between the behavior feature vectors is used to obtain the time-series change feature vector of the trained old man's behavior composed of multiple correlation degrees, where N is the number of the trained old man's behavior feature vector in the sequence of the trained old man's behavior feature vector; Pass the time-series change feature vector of the trained elderly person's behavior through the classifier to obtain a classification loss function value; and use the classification loss function value to compare the behavioral feature extractor based on the convolutional neural network model and the classifier Training is performed, wherein in each iteration of the training, training optimization is performed on the temporal change feature vector of the trained elderly person's behavior.

In the above technical solution, each training old man behavior feature vector in the sequence of training old man behavior feature vectors represents the image semantic feature of the corresponding clipped behavioral key frame. Here, considering that the training behavior collection video is processed In the case of discrete sampling processing to obtain a sequence of training behavior monitoring key frames, and randomly trimming the sequence of training behavior monitoring key frames to obtain a sequence of trimmed behavior key frames, although the temporal direction of the source image semantics can be improved Express randomness, thereby improving the generalization performance of feature regression, but it also significantly increases the uneven distribution of image semantic features caused by the difference in the random distribution of source image semantics, making the top N of the sequence of the training elderly behavior feature vectors separately calculated -The correlation degree between the 1 training old man's behavior feature vector and the Nth training old man's behavior feature vector. The training old man's behavior temporal change feature vector composed of multiple correlation degrees will also have a more significant overall characteristic distribution. Inconsistent and unstable, thus affecting the stability of its classification training through the classifier.

Based on this, the applicant of this application performs training and optimization on the temporal change feature vector of the trained elderly person's behavior in each iteration when classifying and training the temporal change feature vector of the trained elderly person's behavior through a classifier, which is specifically expressed as: In each iteration of the training, the following optimization formula is used to perform training optimization on the behavioral temporal change feature vector of the trained elderly person to obtain the optimal trained elderly behavioral temporal variation feature vector; wherein, the optimization formula is:

;

in, is the time-series change feature vector of the trained elderly person’s behavior/> eigenvalues,/> and/> They are respectively the time series change feature vectors of the trained elderly people’s behavior/> The 1 norm and 2 norm of ,/> is the time-series change feature vector of the trained elderly person’s behavior/> The length of , and/> Yes and/> Relevant weight hyperparameters, /> is the eigenvalue of the time-series change eigenvector of the elderly’s behavior in the optimized training,/> Represents the calculation of the value of a natural exponential function raised to a numerical power.

Here, through the training of the elderly behavior temporal change feature vector The overall feature distribution is represented by the structural consistency and stability under the rigid structure of absolute distance and the non-rigid structure of spatial distance respectively, so as to make the time-series change feature vector of the behavior of the trained elderly/> The global feature distribution has a certain degree of repeatability for local pattern changes, so that the feature vector changes in the time series of the trained elderly person's behavior/> When classifying through a classifier, it is robust to changes in the scale and rotation of the global feature distribution through the weight matrix of the classifier, improving the stability of classification training.

In summary, the wireless anti-fall alarm method for nursing care based on the embodiment of the present invention has been clarified. When it detects that the elderly person's behavior may pose a risk of falling, it will promptly generate an early warning prompt through the calling device to notify relevant personnel for rescue or intervention.

Figure 3 is a block diagram of a wireless anti-fall alarm system for nursing provided in an embodiment of the present invention. As shown in Figure 3, the wireless anti-fall alarm system 200 for nursing includes: a surveillance video collection module 210, used to obtain the behavior surveillance video of the monitored elderly subject collected by the camera; a video transmission module 220, used to The behavior monitoring video is transmitted to the background anti-fall server through the wireless transmission module; the analysis and processing module 230 is used in the background anti-fall server to analyze and process the behavior monitoring video to determine whether to generate an anti-fall alarm. Prompt; and, the early warning prompt generation module 240 is used to generate an early warning prompt by calling the device in response to generating an anti-fall alarm prompt.

Those skilled in the art can understand that the specific operations of each step in the wireless anti-fall alarm system for nursing care have been described in detail in the description of the wireless anti-fall alarm method for nursing care in Figures 1 and 2 above. And therefore, repeated description thereof will be omitted.

As mentioned above, the wireless anti-fall alarm system 200 for nursing care according to the embodiment of the present invention can be implemented in various terminal devices, such as a server for wireless anti-fall alarm for nursing care, etc. In one example, the wireless anti-fall alarm system 200 for nursing care according to an embodiment of the present invention can be integrated into a terminal device as a software module and/or a hardware module. For example, the wireless anti-fall alarm system 200 for nursing can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the wireless anti-fall alarm system 200 for nursing The anti-fall alarm system 200 can also be one of many hardware modules of the terminal device.

Alternatively, in another example, the wireless anti-fall alarm system 200 for nursing and the terminal device may also be separate devices, and the wireless anti-fall alarm system 200 for nursing may be connected via wired and/or wireless devices. The network is connected to the terminal device and transmits interactive information according to the agreed data format.

Figure 4 is an application scenario diagram of a wireless anti-fall alarm method for nursing provided in an embodiment of the present invention. As shown in Figure 4, in this application scenario, first, obtain the behavior monitoring video (for example, C shown in Figure 4) of the monitored elderly subject (for example, M shown in Figure 4) collected by the camera. ); then, input the acquired behavior monitoring video into a server (for example, S as shown in Figure 4) deployed with a wireless anti-fall alarm algorithm for nursing, where the server can be based on the wireless anti-fall alarm algorithm for nursing. The fall alarm algorithm processes the behavioral monitoring video to generate an early warning prompt through the calling device.

The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (8)

1. A wireless anti-fall alarm method for nursing, which is characterized in that it includes: Obtain behavioral monitoring videos of elderly subjects being monitored collected by cameras; Transmit the behavior monitoring video to the backend anti-fall server through the wireless transmission module; In the background anti-fall server, the behavior monitoring video is analyzed and processed to determine whether an anti-fall alarm prompt is generated; In response to generating an anti-fall alarm prompt, an early warning prompt is generated through the calling device; Among them, in the background anti-fall server, the behavior monitoring video is analyzed and processed to determine whether an anti-fall alarm prompt is generated, including: Perform data preprocessing on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames; Extract the timing characteristics of the elderly behavior in the sequence of clipped behavioral key frames to obtain the timing change feature vector of the behavior of the elderly; Based on the time-series change feature vector of the elderly person's behavior, determine whether to generate an anti-fall alarm prompt; Wherein, data preprocessing is performed on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames, including: Perform discrete sampling processing on the behavior monitoring video to obtain a sequence of behavior monitoring key frames; The sequence of behavioral monitoring key frames is randomly trimmed to obtain the sequence of trimmed behavioral key frames. 2. The wireless anti-fall alarm method for nursing according to claim 1, characterized in that extracting the elderly behavior timing characteristics in the sequence of clipped behavioral key frames to obtain the elderly behavior timing change feature vector includes: Using a deep learning network model to extract the behavioral features of the elderly in the sequence of clipped behavioral key frames to obtain a sequence of behavioral feature vectors of the elderly; Calculate the correlation between the first N-1 elderly behavior feature vectors and the Nth elderly behavior feature vector in the sequence of the elderly's behavior feature vectors to obtain the time-series change feature vector of the elderly's behavior composed of multiple correlations. . 3. The wireless anti-fall alarm method for nursing according to claim 2, characterized in that the deep learning network model is a behavioral feature extractor based on a convolutional neural network model; Wherein, the behavioral feature extractor based on the convolutional neural network model includes: input layer, convolution layer, pooling layer, activation layer and output layer. 4. The wireless anti-fall alarm method for nursing according to claim 3, characterized in that a deep learning network model is used to extract the behavioral characteristics of the elderly in the sequence of the clipped behavioral key frames to obtain the behavioral characteristic vector of the elderly. Sequences, including: The sequence of clipped behavioral key frames is passed through the behavioral feature extractor based on the convolutional neural network model to obtain a sequence of behavioral feature vectors of the elderly. 5. The wireless anti-fall alarm method for nursing according to claim 4, wherein N is the number of the elderly behavior feature vectors in the sequence of the elderly behavior feature vectors. 6. The wireless anti-fall alarm method for nursing according to claim 5, characterized in that, based on the timing change feature vector of the elderly person's behavior, determining whether to generate an anti-fall alarm prompt includes: The time-series change feature vector of the elderly person's behavior is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether an anti-fall alarm prompt is generated. 7. The wireless anti-fall alarm method for nursing according to claim 6, further comprising a training step: training the behavioral feature extractor and the classifier based on the convolutional neural network model; Wherein, the training steps include: Obtain training data, which includes training behavior monitoring videos and the true value of whether an anti-fall alarm prompt is generated; In the background anti-fall server, the training behavior monitoring video is discretized and sampled to obtain a sequence of training behavior monitoring key frames; Randomly trim the sequence of training behavior monitoring key frames to obtain a sequence of behavioral key frames after training trimming; Pass the sequence of behavioral key frames after training through the behavioral feature extractor based on the convolutional neural network model to obtain a sequence of trained elderly behavioral feature vectors; Calculate the correlation between the first N-1 behavioral feature vectors of the trained old man and the Nth trained old man's behavior feature vector in the sequence of the trained old man's behavior feature vectors to obtain the time series change of the trained old man's behavior composed of multiple correlation degrees. Feature vector, where N is the number of the training old man's behavior feature vectors in the sequence of the training old man's behavior feature vector; Pass the time-series change feature vector of the trained elderly person's behavior through the classifier to obtain the classification loss function value; The behavioral feature extractor and the classifier based on the convolutional neural network model are trained with the classification loss function value, wherein in each iteration of the training, the temporal changes in the behavior of the trained elderly are Feature vectors are trained and optimized. 8. A wireless anti-fall alarm system for nursing, which is characterized by including: Surveillance video collection module, used to obtain behavioral surveillance videos of elderly subjects being monitored collected by cameras; A video transmission module, used to transmit the behavior monitoring video to the background anti-fall server through the wireless transmission module; An analysis and processing module, used in the background anti-fall server to analyze and process the behavior monitoring video to determine whether an anti-fall alarm prompt is generated; An early warning prompt generation module is used to generate an early warning prompt through the calling device in response to generating an anti-fall alarm prompt; Among them, the early warning prompt generation module includes: Perform data preprocessing on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames; Extract the timing characteristics of the elderly behavior in the sequence of clipped behavioral key frames to obtain the timing change feature vector of the behavior of the elderly; Based on the time-series change feature vector of the elderly person's behavior, determine whether to generate an anti-fall alarm prompt; Wherein, data preprocessing is performed on the behavioral monitoring video to obtain a sequence of trimmed behavioral key frames, including: Perform discrete sampling processing on the behavior monitoring video to obtain a sequence of behavior monitoring key frames; The sequence of behavioral monitoring key frames is randomly trimmed to obtain the sequence of trimmed behavioral key frames.