CN-116548954-B - Fall detection method and device through gravity center change monitoring

CN116548954BCN 116548954 BCN116548954 BCN 116548954BCN-116548954-B

Abstract

The invention relates to the technical field of fall detection, in particular to a fall detection method and device through gravity center change monitoring. A fall detection method through gravity center change monitoring is to collect plantar gravity distribution data in advance and add a state label, input the obtained data into an LSTM neural network, train the mature LSTM neural network, and when a user inputs plantar gravity distribution data again, obtain a prediction result of an action state directly after passing through the mature LSTM neural network. According to the invention, a plurality of pressure sensors are arranged on the sole of a foot, barycentric coordinates are established according to the positions of the pressure sensors and the forces exerted on the pressure sensors, the barycentric coordinates at different moments are input into the LSTM neural network, the LSTM neural network can be trained to form a universal neural network model, and when a user moves in daily, gravity distribution data are input into the LSTM neural network, so that the probability of different postures of the user can be judged to determine whether the user falls.

Inventors

LIAN YANGLIN
DAI HOUDE
XIA XUKE
LIN ZHIRONG
WU LINGYU

Assignees

泉州装备制造研究所

Dates

Publication Date: 20260505
Application Date: 20230614

Claims (9)

1. A fall detection method by gravity center change monitoring, comprising the steps of: S1, acquiring acquisition plantar gravity distribution data based on time sequence data acquired in real time by a plantar pressure sensor array, adding a state label, and calculating barycenter change rates and barycenter change angles at different moments, wherein the barycenter change angles are used for representing barycenter moving directions, and the barycenter change angle calculation formula is as follows: (3) Can obtain the course of action Changing the gravity center of the moment; s2, inputting multidimensional time sequence characteristic data comprising plantar pressure distribution, barycentric coordinates, barycentric change rate and barycentric transformation angle obtained in the step S1 into an LSTM neural network, training the LSTM neural network, and determining model parameters in the LSTM neural network to form a mature LSTM neural network; S3, inputting plantar gravity distribution data, obtaining probability values corresponding to different state labels through a mature LSTM neural network, and selecting a state with the highest probability value as a prediction result, wherein the state labels are divided into normal walking and falling, and the falling state labels are specifically subdivided into forward falling, backward falling, leftward falling and rightward falling; The gravity center coordinates are established for the gravity distribution data collected in the step S1 and the step S3: (1) Wherein, the In order to be the coordinates of the pressure sensor, In response to the pressure experienced by the pressure sensor, To calculate the coordinates of the center of gravity.
2. The fall detection method by gravity center change monitoring according to claim 1, wherein the S1 specific operation steps are: s11, data acquisition, namely, wearing insoles with pressure detection devices by a plurality of subjects to perform normal walking and intentional falling actions, collecting foot gravity distribution conditions of the subjects in different states according to the positions of the pressure sensors to generate corresponding pressure data ; S12, data set making and label adding, namely calculating coordinates of barycenters at different moments according to the formula (1) Establishing a corresponding rectangular coordinate system, and adding a state label to the generated pressure data due to the known action state of the subject; s13, calculating the gravity center change rate and the gravity center change angle, namely obtaining the gravity center coordinate at a moment, and then calculating a formula according to the following gravity center change rate: (2) Can obtain the course of action Rate of change of center of gravity between moments.
3. The fall detection method according to claim 2, wherein in the step S11, the pressure sensor arrays are distributed in the insole, and are respectively set as follows according to the positions of the pressure sensors 、 ··· The pressure data generated corresponding to the pressure sensor position is 、 ··· 。
4. A fall detection method as claimed in claim 3, wherein the data input in step S2 is [ ] ··· , , , , ]。
5. A fall detection method as claimed in claim 4, wherein the LSTM neural network involved in step S2 has the following structure: the first layer is the input layer, and the input variable is [ [ ··· , , , , ]; The second layer is an intermediate layer, an LSTM network model is arranged in the intermediate layer, and the obtained prediction data is processed and then is output to the output layer; The third layer is an output layer, after receiving data from the middle layer, the probabilities of four falling directions are obtained through a softmax function, and the direction with the highest probability is selected as a prediction result.
6. The fall detection method according to claim 5, wherein the intermediate layer includes an LSTM network model, a state layer, a fully connected layer FC, and Relu activation functions, the input feature value is passed through the LSTM neural network model to obtain the state layer of the state, and then is input into the fully connected layer FC, and the fully connected layer FC is provided with the Relu activation function, so as to enhance the nonlinear capability of the LSTM network model.
7. The fall detection method according to claim 1, wherein the gravity distribution data of the sole input in the step S3 is data generated by a user, the subject in the step S1 performs a corresponding action, thereby generating a mature LSTM neural network, and when the user inputs the gravity distribution data of the sole again in the step S3, the probability of different states can be determined, and the state with the highest probability is selected as the prediction result.
8. A fall detection device that monitors through a change in center of gravity, comprising: a user device which is an insole with a pressure detection device; the pressure sensors are distributed in the insole corresponding to the arrays of different areas of the sole and are used for detecting the gravity distribution of the foot; A processing unit configured to perform the method of any of claims 1-7: A storage unit for receiving and storing data from the pressure sensor; a processing unit for performing calculation processing on the data stored in the storage unit; And the neural network model part is used for receiving the data of the processing part and outputting a state prediction result.
9. A fall detection apparatus as claimed in claim 8, further comprising an alarm device coupled to the processing unit for generating an alarm signal when the output is a fall.

Description

Fall detection method and device through gravity center change monitoring Technical Field The invention relates to the technical field of fall detection, in particular to a fall detection method and device monitored by gravity center change. Background Along with rapid development of science and technology, the pace of life of people is faster and faster, the attention to healthy life is higher and higher, and along with the extension of life span, population aging becomes a prominent problem in society, and for the elderly, the damage caused by falling is serious, and even the life of the elderly may be endangered. Aiming at the situation, various fall detection and alarm products in the market are developed and developed vigorously, and various fall detection and alarm products appear. In the fall detection and alarm products in the prior art, the basic principle of the work is that the detection value of one or more physical or movement characteristics of a user wearing the fall detection and alarm product is detected by a sensor, the physical state of the user is deduced according to the detection value, whether the user falls is judged, and alarm processing and the like are performed when the user judges that the user falls. The judgment method is effective directly and has the most wide practical application range, but has the disadvantage that the threshold value for judging the physical state is too direct, and is often simply defined as that the falling angle exceeds a certain value or the acceleration exceeds a certain value, and the specific falling direction cannot be judged, but the action mode and the specific posture of each person are different, and the falling is easily caused by a judgment method of only exceeding a certain threshold value without supporting a large data amount. Disclosure of Invention Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and the appended drawings. The invention aims to overcome the defects, and provides a fall detection method and a fall detection device through gravity center change monitoring, wherein a plurality of pressure sensors are arranged on the sole of a foot, gravity center coordinates are established according to the positions of the pressure sensors and the forces born by the pressure sensors, after the gravity center coordinates at different moments are changed and input into an LSTM neural network, the LSTM neural network can be trained to form a universal neural network model, and when a user inputs gravity distribution data into the LSTM neural network in the daily action process, the probability of different postures of the user can be judged to determine whether the user falls. The invention provides a fall detection method through gravity center change monitoring, which comprises the following steps: S1, collecting plantar gravity distribution data, adding a state label, and calculating barycenter change rates and barycenter change angles at different moments; S2, inputting the data obtained in the S1 into an LSTM neural network, training the LSTM neural network, and determining model parameters in the LSTM neural network to form a mature LSTM neural network; s3, inputting plantar gravity distribution data, obtaining probability values corresponding to different state labels through a mature LSTM neural network, and selecting a state with the highest probability value as a prediction result; The gravity center coordinates are established for the gravity distribution data collected in the step S1 and the step S3: Where (a i,bi) is the pressure sensor coordinate, F i is the pressure applied by the pressure sensor, and (x, y) is the coordinate of the calculated center of gravity. In some embodiments, the specific operation steps of S1 are: S11, data acquisition, namely, wearing insoles with pressure detection devices by a plurality of subjects to perform normal walking and intentional falling actions, collecting foot gravity distribution conditions of the subjects in different states according to the positions of the pressure sensors to generate Corresponding pressure data F i; S12, data set making and adding labels, namely calculating coordinates (x, y) of the centers of gravity at different moments according to a formula (1) and establishing a corresponding rectangular coordinate system, wherein the state labels can be added to the generated pressure data due to the known action state of the subject, and the method comprises the following steps of Status labels can be classified into normal walking and falling; s13, calculating the gravity center change rate and the gravity center change angle, namely obtaining the gravity center coordinate at