A fall detection method and apparatus by monitoring of center of gravity changes

Abstract

This invention relates to the field of fall detection technology, and more particularly to a fall detection method and apparatus based on monitoring changes in the center of gravity. The fall detection method, based on monitoring changes in the center of gravity, involves pre-collecting foot weight distribution data and adding state labels. The resulting data is then input into an LSTM neural network, thereby training a mature LSTM neural network. When the user inputs foot weight distribution data again, the mature LSTM neural network can directly derive a prediction of the user's movement state. This invention places multiple pressure sensors on the sole of the foot and establishes a center of gravity coordinate system based on the sensor positions and the forces acting on them. The changes in the center of gravity coordinates at different times are input into the LSTM neural network, which is then trained to form a general-purpose neural network model. When the user performs daily activities, the weight distribution data is input into this LSTM neural network, which can determine the probability of different postures to ascertain whether a fall has occurred.

Description

Technical Field

[0001] This invention relates to the field of fall detection technology, and in particular to a fall detection method and apparatus that monitors changes in the center of gravity. Background Technology

[0002] With the rapid development of technology, people's pace of life is accelerating, and their focus on healthy living is increasing. Furthermore, with the increase in life expectancy, population aging has become a prominent social issue. For the elderly, injuries from falls are particularly severe and can even be life-threatening. In response to this situation, the research and development of various fall detection and alarm products is booming, resulting in a wide variety of fall detection and alarm products on the market.

[0003] The basic working principle of existing fall detection and alarm products is to detect one or more body or movement characteristics of the user wearing the fall detection and alarm product through sensors, infer the user's physical state based on the detection values, and then determine whether a fall has occurred, and to trigger an alarm when a fall is determined to have occurred.

[0004] This method of judgment is direct and effective, and has the widest range of practical applications. However, its drawback is that the threshold used to judge the body's condition is too direct. It is often simply defined as a fall if the tilt angle exceeds a certain value or the acceleration exceeds a certain value. Furthermore, it cannot determine the specific direction of the fall. However, everyone's behavior and specific posture are different. Without the support of a large amount of data, the method of judging a fall by exceeding a certain threshold is prone to errors and biases. Summary of the Invention

[0005] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description and other accompanying drawings.

[0006] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a fall detection method and device that monitors changes in the center of gravity. Multiple pressure sensors are installed on the soles of the feet, and the center of gravity coordinates are established based on the positions of the pressure sensors and the forces they experience. After inputting the changes in the center of gravity coordinates at different times into an LSTM neural network, the LSTM neural network can be trained to form a general-purpose neural network model. When the user is in daily activities, gravity distribution data is input into the LSTM neural network, and the probability of different postures can be determined to determine whether a fall has occurred.

[0007] This invention provides a fall detection method based on monitoring changes in the center of gravity, comprising the following steps:

[0008] S1: Collect foot gravity distribution data and add status labels, calculate the rate of change of center of gravity and the angle of change of center of gravity at different times;

[0009] S2: Input the data obtained in S1 into the LSTM neural network, train the LSTM neural network, determine the model parameters in the LSTM neural network, and form a mature LSTM neural network.

[0010] S3: Input foot gravity distribution data, and after passing through a mature LSTM neural network, obtain the probability values ​​corresponding to different state labels, and select the state with the highest probability value as the prediction result.

[0011] Specifically, the centroid coordinates are established for the gravity distribution data collected in steps S1 and S3:

[0012]

[0013] Among them, (a i ,b i ) represents the coordinates of the pressure sensor, F i The pressure is the pressure applied to the pressure sensor, and (x,y) are the calculated coordinates of the center of gravity.

[0014] In some embodiments, the specific operation steps of S1 are as follows:

[0015] S11: Data Acquisition: Multiple subjects wore insoles equipped with pressure detection devices and performed normal walking and intentional falls. Based on the location of the pressure sensors, the distribution of foot gravity in different states was collected, generating...

[0016] Corresponding pressure data F i ;

[0017] S12: Dataset Creation and Labeling: The coordinates (x, y) of the center of gravity at different times are calculated according to equation (1), and a corresponding rectangular coordinate system is established. Since the subject's action state is known, state labels can be added to the generated pressure data.

[0018] Status labels can be divided into normal walking and falling;

[0019] S13: Calculate the rate of change of the center of gravity and the angle of change of the center of gravity: After obtaining the coordinates of the center of gravity at a certain moment, based on the following...

[0020] Formula for calculating the rate of change:

[0021]

[0022] The rate of change of the center of gravity between time t2 and t1 during the action can be obtained;

[0023] The following formula is used to calculate the angle of change of the center of gravity:

[0024]

[0025] The change in the center of gravity between time t2 and t1 during the action can be derived.

[0026] In some embodiments, in step S11, the pressure sensor array is distributed in the insole, and is respectively set as (a1, b1), (a2, b2) ... (a...b1) according to the position of the pressure sensor. 13 ,b 13 The pressure data generated at the corresponding pressure sensor location are F1, F2, ..., F. 13 During a gait cycle, the subject's center of gravity changes continuously, gradually shifting from the heel to the forefoot, resulting in different rates of change and angles of change of the center of gravity under different conditions.

[0027] In some embodiments, in step S12, the fall status label can be further subdivided into falling forward, falling backward, falling to the left, and falling to the right.

[0028] In some embodiments, the LSTM neural network involved in step S2 has the following structure:

[0029] The first layer is the input layer, and the input variables are [F1···F 13 ,x,y,v x ,φ];

[0030] The second layer is the intermediate layer, which contains an LSTM network model. The resulting prediction data is processed and then sent to the output layer.

[0031] The third layer is the output layer. After receiving data from the intermediate layer, it uses the softmax function to obtain the probabilities of the four falling directions and selects the direction with the highest probability as the prediction result.

[0032] In some embodiments, the intermediate layer includes an LSTM network model, a STATES layer, a fully connected FC layer, and a ReLU activation function. The input feature values ​​are passed through the LSTM neural network model to obtain the STATES layer, and then input into the fully connected FC layer. The ReLU activation function is set after the fully connected FC layer to enhance the nonlinearity of the LSTM network model.

[0033] In some embodiments, the plantar gravity distribution data input in step S3 is data generated by the user. The subject in step S1 performs a corresponding action, thereby generating a mature LSTM neural network. Then, in step S3, the user inputs the plantar gravity distribution data again, which can determine the probability of different states and select the state with the highest probability as the prediction result.

[0034] The present invention also provides a fall detection device that monitors changes in center of gravity, including a user device, several pressure sensors, and a processing unit. The user device is an insole with a pressure detection device. The pressure sensors are arrayed in the insole corresponding to different areas of the sole of the foot and are used to detect the distribution of foot weight. The processing unit is configured as follows: a storage unit for receiving and storing data from the pressure sensors; a processing unit for performing calculations on the data stored in the storage unit; and a neural network model unit for receiving data from the processing unit and outputting state prediction results.

[0035] In some embodiments, the fall detection device further includes an alarm device connected to the processing unit, which issues an alarm signal when the output result indicates a fall.

[0036] By adopting the above technical solution, the beneficial effects of the present invention are:

[0037] This invention incorporates multiple pressure sensors within the insole and establishes a center of gravity coordinate system based on the sensor positions and the forces acting upon them. By inputting the changes in the center of gravity coordinates at different times into an LSTM neural network, the network can be trained to form a general-purpose neural network model. When a user inputs gravity distribution data into the LSTM neural network during daily activities, the system can determine the probability of different postures to identify whether a fall has occurred and issue an appropriate alarm upon detecting a fall.

[0038] This invention introduces gravity coordinates and an LSTM neural network into the conventional method of determining whether a fall has occurred using pressure sensors. Gravity coordinates can calculate the rate of change of the center of gravity and the coordinates of the center of gravity throughout the entire movement process by using the position of the pressure sensor. The LSTM neural network model has the characteristic of storing a large amount of information over a long period of time. After a mature LSTM neural network is built, the neural network can determine the posture of the user through a long period of behavior, rather than judging whether the user has fallen based on a single threshold.

[0039] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0040] Undoubtedly, such and other objects of the present invention will become more apparent after the following detailed description of the preferred embodiments, which are illustrated in various accompanying drawings and figures.

[0041] To make the above and other objects, features and advantages of the present invention more apparent and understandable, one or more preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0042] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0043] In the accompanying drawings, the same parts use the same reference numerals, and the drawings are schematic and not necessarily drawn to actual scale.

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only one or more embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on such drawings without creative effort.

[0045] Figure 1 This is a schematic diagram illustrating the establishment of a mature LSTM neural network in some embodiments of the present invention;

[0046] Figure 2 This is a schematic diagram showing the distribution of pressure sensors and the center of gravity change curves within one gait cycle in some embodiments of the present invention.

[0047] Figure 3 This is a schematic diagram showing the rate of change of the center of gravity and the angle of change of the center of gravity under different states in some embodiments of the present invention;

[0048] Figure 4 This is a schematic diagram of the neural network determining the action state in some embodiments of the present invention;

[0049] Figure 5 This is a schematic diagram of the insole structure in some embodiments of the present invention;

[0050] Figure 6 This is a schematic diagram of the LSTM neural network structure in some embodiments of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0052] Furthermore, in the description of this invention, it should be understood that the terms "center," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0053] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral unit; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. However, specifying a direct connection indicates that the two main bodies are not connected through a transitional structure, but rather formed as a whole through a connecting structure. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0054] In this invention, unless otherwise expressly specified and limited, the first feature "on" or "below" the second feature may be in direct contact with the first and second features, or indirect contact through an intermediate medium. In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0055] Reference Figure 1-6 , Figure 1 This is a schematic diagram illustrating the establishment of a mature LSTM neural network in some embodiments of the present invention; Figure 2 This is a schematic diagram showing the distribution of pressure sensors and the center of gravity change curves within one gait cycle in some embodiments of the present invention. Figure 3 This is a schematic diagram showing the rate of change of the center of gravity and the angle of change of the center of gravity under different states in some embodiments of the present invention; Figure 4 This is a schematic diagram of the neural network determining the action state in some embodiments of the present invention; Figure 5 This is a schematic diagram of the insole structure in some embodiments of the present invention; Figure 6 This is a schematic diagram of the LSTM neural network structure in some embodiments of the present invention.

[0056] like Figure 1 As shown, a fall detection method that monitors changes in center of gravity includes the following steps:

[0057] S1: Collect foot gravity distribution data and add status labels, calculate the rate of change of center of gravity and the angle of center of gravity transformation at different times:

[0058] The specific operating steps for S1 are as follows:

[0059] S11: Data Acquisition: Multiple subjects wore insoles equipped with pressure detection devices and performed normal walking and intentional falls. Based on the location of the pressure sensor, the distribution of foot gravity in different states was collected, generating corresponding pressure data F. i ;

[0060] like Figure 2-3 As shown, the pressure sensor array is distributed in the insole, and according to the position of the pressure sensor, it is designated as (a1, b1), (a2, b2), ..., (a...). 13 ,b 13 The pressure data generated at the corresponding pressure sensor location are F1, F2, ..., F. 13 During a gait cycle, the subject's center of gravity changes continuously, gradually shifting from the heel to the forefoot, resulting in different rates of change and angles of change of the center of gravity under different conditions.

[0061] S12: Dataset creation and labeling: The coordinates (x, y) of the center of gravity at different times are calculated according to formula (1) and a corresponding rectangular coordinate system is established. Since the subject's movement state is known, a state label can be added to the generated pressure data. The state label can be divided into normal walking and falling.

[0062] like Figure 4 As shown, the fall status label can be further divided into falling forward, falling backward, falling to the left, and falling to the right.

[0063] S13: Calculate the rate of change of the center of gravity and the angle of change of the center of gravity: After obtaining the coordinates of the center of gravity at a certain moment, calculate the rate of change of the center of gravity according to the following formula:

[0064]

[0065] The rate of change of the center of gravity between time t2 and t1 during the action can be obtained;

[0066] The following formula is used to calculate the angle of change of the center of gravity:

[0067]

[0068] The change in the center of gravity between time t2 and t1 during the action can be derived.

[0069] S2: Input the data obtained in S1 into the LSTM neural network, train the LSTM neural network, determine the model parameters in the LSTM neural network, and form a mature LSTM neural network.

[0070] like Figure 6 As shown, the LSTM neural network involved in step S2 has the following structure:

[0071] The first layer is the input layer, and the input variables are [F1···F 13 ,x,y,v x ,φ];

[0072] The second layer is the intermediate layer, which contains an LSTM network model. The resulting prediction data is processed and then sent to the output layer.

[0073] The third layer is the output layer. After receiving data from the intermediate layer, it uses the softmax function to obtain the probabilities of the four falling directions and selects the direction with the highest probability as the prediction result.

[0074] Specifically, the intermediate layer includes an LSTM network model, a STATES layer, a fully connected FC layer, and a ReLU activation function. The input feature values ​​are processed by the LSTM neural network model to obtain the STATES layer, and then input into the fully connected FC layer. The ReLU activation function is set after the fully connected FC layer to enhance the nonlinearity of the LSTM network model.

[0075] S3: Input foot gravity distribution data, and after passing through a mature LSTM neural network, obtain the probability values ​​corresponding to different state labels, and select the state with the highest probability value as the prediction result.

[0076] Specifically, the centroid coordinates are established for the gravity distribution data collected in steps S1 and S3:

[0077]

[0078] Among them, (a i ,b i ) represents the coordinates of the pressure sensor, F i The pressure is the pressure applied to the pressure sensor, and (x,y) are the calculated coordinates of the center of gravity.

[0079] The foot gravity distribution data input in step S3 is user-generated data. The subject in step S1 performs a corresponding action, thereby generating a mature LSTM neural network. When the user inputs foot gravity distribution data again, the probability of different states can be determined, and the state with the highest probability is selected as the prediction result.

[0080] like Figure 5 As shown, the present invention also provides a fall detection device that monitors changes in center of gravity, including a user device, several pressure sensors, and a processing unit. The user device is an insole with a pressure detection device. The pressure sensors are disposed in the insole corresponding to different areas of the sole of the foot to detect the distribution of foot weight. The processing unit is configured as follows: a storage unit for receiving and storing data from the pressure sensors; a processing unit for calculating and processing the data stored in the storage unit; and a neural network model unit for receiving data from the processing unit and outputting state prediction results.

[0081] Optionally, according to some embodiments of the present invention, the fall detection device further includes an alarm device connected to the processing unit, which issues an alarm signal when the output result indicates a fall.

[0082] This invention establishes a center of gravity coordinate system by placing multiple pressure sensors on the soles of the feet and using the positions of these sensors and the forces they experience. The changes in the center of gravity coordinates at different times are then input into an LSTM neural network, which is trained to form a general-purpose neural network model. During daily activities, gravity distribution data is input into this mature LSTM neural network, which can determine the probability of the user being in different postures and select the posture with the highest probability as the prediction result. Specifically, the fall can be classified as a forward fall, a backward fall, a left fall, or a right fall. After a fall is determined, a signal can be transmitted to an alarm device to alert others that the user is about to fall or has already fallen.

[0083] It should be understood that the embodiments disclosed herein are not limited to the specific processing steps or materials disclosed herein, but should be extended to equivalent substitutions of such features as understood by those skilled in the art. It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0084] The term "embodiment" in this specification refers to a specific feature or characteristic described in connection with an embodiment that is included in at least one embodiment of the invention. Therefore, phrases or "embodiments" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0085] Furthermore, the described features or characteristics can be incorporated into one or more embodiments in any other suitable manner. In the above description, specific details, such as thickness, quantity, etc., are provided to provide a comprehensive understanding of embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented without the aforementioned specific details or may be implemented using other methods, components, materials, etc.

Claims

1. A fall detection method based on monitoring changes in center of gravity, characterized in that, Includes the following steps: S1: Based on real-time data collected by the plantar pressure sensor array, acquire plantar gravity distribution data and add status tags. Calculate the rate of change of the center of gravity and the angle of change of the center of gravity at different times. The angle of change of the center of gravity is used to characterize the direction of movement of the center of gravity. The calculation formula for the angle of change of the center of gravity is as follows: (Equation 3) It can be concluded that during the action The shift of the center of gravity over time; S2: Input the multi-dimensional time-series feature data obtained in S1, including plantar pressure distribution, center of gravity coordinates, rate of change of center of gravity and angle of change of center of gravity, into the LSTM neural network, train the LSTM neural network, determine the model parameters in the LSTM neural network, and form a mature LSTM neural network. S3: Input foot gravity distribution data, and after passing through a mature LSTM neural network, the probability values ​​corresponding to different state labels can be obtained. The state with the highest probability value is selected as the prediction result. Among them, the state label is divided into normal walking and falling. The falling state label is further subdivided into falling forward, falling backward, falling to the left, and falling to the right. Specifically, the centroid coordinates are established for the gravity distribution data collected in steps S1 and S3: (Equation 1) in, For the coordinates of the pressure sensor, The pressure applied to the pressure sensor. These are the coordinates of the calculated center of gravity.

2. The fall detection method based on monitoring changes in center of gravity according to claim 1, characterized in that, The specific operating steps for S1 are as follows: S11: Data Acquisition: Multiple subjects wore insoles equipped with pressure detection devices and performed normal walking and intentional falls. Based on the location of the pressure sensor, the distribution of foot gravity in the subjects' feet under different conditions was collected, generating corresponding pressure data. ; S12: Dataset Creation and Labeling: The coordinates of the centroid at different times are calculated according to equation (1). A corresponding rectangular coordinate system is established, and since the subject's action state is known, state labels can be added to the generated pressure data; S13: Calculate the rate of change of the center of gravity and the angle of change of the center of gravity: After obtaining the coordinates of the center of gravity at a certain moment, calculate the rate of change of the center of gravity according to the following formula: (Equation 2) It can be concluded that during the action The rate of change of the center of gravity over time.

3. The fall detection method based on monitoring changes in center of gravity according to claim 2, characterized in that, In step S11, the pressure sensor array is distributed within the insole, and is configured according to the location of the pressure sensors. , ··· The pressure data generated at the corresponding pressure sensor location is , ··· .

4. The fall detection method based on monitoring changes in center of gravity according to claim 3, characterized in that, The data input in step S2 is [ ··· , , , , ].

5. The fall detection method based on monitoring changes in center of gravity according to claim 4, characterized in that, The LSTM neural network involved in step S2 has the following structure: The first layer is the input layer, and the input variables are [ ··· , , , , ]; The second layer is the intermediate layer, which contains an LSTM network model. The resulting prediction data is processed and then sent to the output layer. The third layer is the output layer. After receiving data from the intermediate layer, it uses the softmax function to obtain the probabilities of the four falling directions and selects the direction with the highest probability as the prediction result.

6. The fall detection method based on monitoring changes in center of gravity according to claim 5, characterized in that, The intermediate layer includes an LSTM network model, a STATES layer, a fully connected (FC) layer, and a ReLU activation function. The input feature values ​​are processed by the LSTM neural network model to obtain the STATES layer, and then input into the fully connected (FC) layer. The FC layer is followed by the ReLU activation function to enhance the nonlinearity of the LSTM network model.

7. The fall detection method based on monitoring changes in center of gravity according to claim 1, characterized in that, The foot gravity distribution data input in step S3 is user-generated data. The subject in step S1 performs a corresponding action, thereby generating a mature LSTM neural network. Then, in step S3, the user inputs the foot gravity distribution data again, which can determine the probability of different states and select the state with the highest probability as the prediction result.

8. A fall detection device that monitors changes in center of gravity, characterized in that, include: User equipment, which is an insole with a pressure detection device; Several pressure sensors, corresponding to different areas of the sole of the foot, are arrayed within the insole to detect the distribution of foot weight. A processing unit configured to perform the method as described in any one of claims 1-7: The storage unit is used to receive and store data from the pressure sensor; The processing unit is used to perform calculations and processing on the data stored in the storage unit; The neural network model unit receives data from the processing unit and outputs state prediction results.

9. The fall detection device for monitoring changes in center of gravity according to claim 8, characterized in that, It also includes an alarm device connected to the processing unit, which issues an alarm signal when the output indicates a fall.

Images