CN112669568A - Multi-mode human body falling detection method - Google Patents
Multi-mode human body falling detection method Download PDFInfo
- Publication number
- CN112669568A CN112669568A CN202011505755.9A CN202011505755A CN112669568A CN 112669568 A CN112669568 A CN 112669568A CN 202011505755 A CN202011505755 A CN 202011505755A CN 112669568 A CN112669568 A CN 112669568A
- Authority
- CN
- China
- Prior art keywords
- detection object
- detection method
- particle
- mode human
- support vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses a multi-mode human body falling detection method, which comprises the following steps: acquiring acceleration data of a detection object; judging the state of the detection object according to the acceleration data; collecting surface electromyographic signals of the detection object when the detection object is judged to be in a falling state; extracting electromyographic signal characteristics from the surface electromyographic signals; identifying the electromyographic signal characteristics through a support vector machine; and triggering an alarm system when the support vector machine identifies that the detected object is in a falling state. The invention provides a multi-mode human body falling detection method, which aims at solving the problems that the existing falling identification method is easily interfered by external noise to cause low identification accuracy and a support vector machine is greatly influenced by a punishment parameter c and a kernel function parameter g.
Description
Technical Field
The invention belongs to the technical field of intelligent old people care of Internet of things and artificial intelligence, and particularly relates to a multi-mode human body falling detection method.
Background
With the aging problem of the population becoming more and more serious, the safety problem of the mind and the body of the elderly people becomes one of the focus problems of social attention. With the increase of age, the bone texture of the old is loose, and walking and falling down become an important factor which damages the body health of the old in daily life. The traditional falling monitoring and identifying method is mainly used for completing the falling monitoring and identifying of the old through equipment such as video images, sound or vibration, plantar pressure sensors and the like, and is simple in identification model, fuzzy in identification process and free of inspection. In recent years, with the rise of wearable equipment, the three-axis acceleration sensing technology and the myoelectric signal acquisition instrument technology are continuously improved, the better monitoring equipment is provided for improving the identification of the falling behavior, and the effectiveness and the reliability of the identification of the falling behavior of the old are effectively guaranteed.
In the existing method, the triaxial acceleration sensing has certain accuracy in judging the falling behavior, but since the three-dimensional angular acceleration reactor is interfered by external noise, the data measurement result is easily misled in the system operation process, and the triaxial acceleration sensing is singly utilized to judge the falling behavior with low accuracy and high monitoring difficulty.
There are many different methods for surface electromyogram signal identification and feature extraction, and common methods such as a neural network method and cluster analysis make certain progress in surface electromyogram signal identification, but all have certain disadvantages. A Support Vector Machine (SVM) (support Vector machine) is a supervised classifier, has a plurality of special advantages in the process of solving the recognition of small samples, nonlinearity and Gaussian patterns, and can solve the problems of nonstandard theoretical guidance of a neural network structure, long learning time, local minimum points and the like, but the algorithm is difficult to determine by a penalty parameter c and a kernel function parameter g.
Disclosure of Invention
The invention provides a multi-mode human body falling detection method, which adopts the following technical scheme:
a multi-mode human fall detection method comprises the following steps:
acquiring acceleration data of a detection object;
judging the state of the detection object according to the acceleration data;
collecting surface electromyographic signals of the detection object when the detection object is judged to be in a falling state;
extracting electromyographic signal characteristics from the surface electromyographic signals;
identifying the electromyographic signal characteristics through a support vector machine;
and triggering an alarm system when the support vector machine identifies that the detected object is in a falling state.
Further, acceleration data of the detection object is acquired through the three-axis acceleration sensor.
Furthermore, the acceleration data of the detection object collected by the three-axis acceleration sensor contains acceleration values a of X, Y and Z axesx,ay,az。
Further, a specific method of determining the state of the detection object from the acceleration data is:
the supination angle Pitch, the rollover angle Roll, and the left-right rotation angle Yaw are calculated according to the following formulas:
and judging whether the detected object falls down or not according to the comparison of the supination angle Pitch, the rollover angle Roll and the left and right rotation angle Yaw with the corresponding normal values.
Further, the supine angle Pitch, the rollover angle Roll and the left and right rotation angle Yaw are optimized through an Alemann filter algorithm.
Further, a specific method for extracting the electromyographic signal features from the electromyographic signal is as follows:
preprocessing the surface myoelectric signal;
extracting electromyographic signal characteristics from the preprocessed surface electromyographic signals through wavelet packet arrangement combination entropy.
Further, the specific method for preprocessing the surface myoelectric signal is as follows:
and extracting information of a 16Hz-160Hz frequency band from the surface electromyogram signal.
Further, a specific method for extracting electromyographic signal characteristics from the preprocessed surface electromyographic signal through wavelet packet permutation entropy includes:
carrying out five-layer wavelet packet decomposition on the surface myoelectric signals;
and calculating the permutation entropy of the surface electromyographic signals after the nine low-frequency subspaces are reconstructed by using a permutation entropy formula to serve as the electromyographic signal characteristics.
Further, a penalty parameter c and a kernel function parameter g of the support vector machine are optimized through a particle swarm optimization PSO.
Further, a specific method for optimizing the penalty parameter c and the kernel function parameter g of the support vector machine by the particle swarm optimization PSO comprises the following steps:
1) by the pair of learning factors c1，c2Optimizing the weight coefficient, initializing the position and the speed of the particles, and initially setting the initial position of each particle as the initial best position;
2) calculating the fitness of each particle, and evaluating the fitness of the particles by adopting K-fold cross validation;
3) updating the speed and the position of the particle according to a speed and position calculation formula of a PSO algorithm;
the speed and position of the PSO algorithm are calculated as follows:
vij(t+1)＝wvij(t)+c1r1[pij-xij(t)]+c2r2[pgj-xij(t)]，
xij(t+1)＝xij(t)+vij(t+1)，
wherein, c1,c1Is a learning factor, r1,r2Is [0, 1 ]]W is a weight coefficient. p is a radical ofijRepresenting the local optimum position, p, of the particle i in the j-th dimensiongjRepresenting the global optimal position of the particle i in the j dimension; v. ofij(t) represents the flight velocity of the particle i in the j-th dimension in the t-th generation. x is the number ofij(t) represents the current position of the particle i in the j dimension in the t generation;
4) and checking whether a termination condition is met, if so, mapping the group optimal particles into the optimal solution of the penalty parameter c and the kernel function parameter g of the support vector machine, otherwise, turning to the step 2 and continuing a new round of search.
The invention has the beneficial effects that the multi-mode human body falling detection method provided by the invention is used for solving the problems that the existing falling identification method is easily interfered by external noise to cause low identification accuracy and the support vector machine is greatly influenced by punishment parameters c and kernel function parameters g, and provides the multi-mode detection method which is integrated with triaxial acceleration induction and PSO (particle swarm optimization) SVM to improve the identification accuracy of the falling behavior of the old.
Drawings
Fig. 1 is a schematic diagram of a multi-mode human fall detection method of the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and the embodiments.
Fig. 1 shows a multi-mode human fall detection method, which includes the following steps: and S1, acquiring acceleration data of the detection object. And S2, judging the state of the detection object according to the acceleration data. And S3, acquiring the surface electromyographic signals of the detection object when the detection object is judged to fall. And S4, extracting electromyographic signal characteristics from the surface electromyographic signals. And S5, identifying the electromyographic signal characteristics through a support vector machine. And S6, triggering an alarm system when the support vector machine identifies that the detected object is in a falling state. According to the method, after the falling behavior is monitored through the timely speed data, the falling behavior is identified and verified again through the support vector machine, and at the moment, the alarm system is triggered. It will be appreciated that to cope with extreme conditions, the system incorporates a manual control system, i.e. the user can manually complete the alarm settings themselves. The above steps are specifically described below.
For step S1, acceleration data of the test object is acquired.
In the invention, the acceleration data of the detection object is acquired by the three-axis acceleration sensor.
Specifically, the acceleration data of the detection object collected by the three-axis acceleration sensor includes acceleration values a of the X, Y, and Z axesx,ay,az。
In step S2, the state of the detection object is determined based on the acceleration data.
Specifically, the specific method for determining the state of the detection object from the acceleration data is as follows:
the supination angle Pitch, the rollover angle Roll, and the left-right rotation angle Yaw are calculated according to the following formulas:
and then, whether the detected object falls down is judged according to the comparison of the supine angle Pitch, the side turning angle Roll and the left and right rotation angle Yaw with the corresponding normal values. Firstly, whether the lying angle Pitch is higher than a normal value or not is monitored, if not, whether the rollover angle Roll and the left and right rotation angle Yaw are higher than the normal value or not is continuously monitored, and if not, a pre-alarm mode is started.
Because the triaxial acceleration sensor can be interfered by external noise, the data measurement result is easily misled in the system operation process, and in the application, the calculated supination angle Pitch, rollover angle Roll and left and right rotation angle Yaw are optimized by adopting an Alemann filtering algorithm.
In step S3, a surface electromyogram signal of the detection object is acquired when it is determined that the detection object is in a falling state.
And when the lying angle Pitch, the side turning angle Roll and the left and right rotation angle Yaw greatly exceed normal values, turning on a switch of the electromyographic signal acquisition instrument, entering an electromyographic acquisition behavior, and acquiring surface electromyographic signals of the user.
For step S4, extracting electromyographic signal features from the surface electromyographic signal.
The specific method for extracting the electromyographic signal features from the electromyographic signals comprises the following steps:
the surface myoelectric signal is pre-processed. Specifically, information of a 16Hz-160Hz frequency band is extracted from the surface electromyogram signal, and high-frequency noise and unnecessary low-frequency information are filtered.
Extracting electromyographic signal characteristics from the preprocessed surface electromyographic signals through wavelet packet arrangement combination entropy.
Specifically, five-layer wavelet packet decomposition is performed on the surface myoelectric signals. And calculating the permutation entropy of the surface electromyographic signals after the nine low-frequency subspaces are reconstructed by using a permutation entropy formula to serve as the electromyographic signal characteristics.
Wherein, the permutation and combination entropy formula is as follows:
wherein the content of the first and second substances,
t is a time sequence; x is the number ofi,xi+1,…,xi+n-1Represents n consecutive sample points in the time series, and the count represents the number of times the permutation condition pi occurs in the sequence.
For step S5, electromyographic signal features are identified by a support vector machine.
In the application, in order to improve the classification accuracy of the SVM, a penalty parameter c and a kernel function parameter g of the SVM are optimized through a Particle Swarm Optimization (PSO).
The specific method for optimizing the penalty parameter c and the kernel function parameter g of the support vector machine by the particle swarm optimization PSO comprises the following steps:
1. by the pair of learning factors c1，c2And optimizing the weight coefficients, initializing the position and the speed of the particles, and initially setting the initial position of each particle as the initial best position.
2. And calculating the fitness of each particle, wherein the fitness of the particles is evaluated by adopting K-fold cross validation.
3. And updating the speed and the position of the particle according to a speed and position calculation formula of the PSO algorithm.
The speed and position of the PSO algorithm are calculated as follows:
vij(t+1)＝wvij(t)+c1r1[pij-xij(t)]+c2r2[pgj-xij(t)]，
xij(t+1)＝xij(t)+vij(t+1)，
wherein, c1,c1Is a learning factor, r1,r2Is [0, 1 ]]W is a weight coefficient. p is a radical ofijRepresenting the local optimum position, p, of the particle i in the j-th dimensiongjRepresenting the global optimal position of the particle i in the j-th dimension. v. ofij(t) represents the flight velocity of the particle i in the j-th dimension in the t-th generation. x is the number ofij(t) represents the current position of the particle i in the j-th dimension in the t-th generation.
4. And checking whether a termination condition is met, if so, mapping the group optimal particles into the optimal solution of the penalty parameter c and the kernel function parameter g of the support vector machine, otherwise, turning to the step 2 and continuing a new round of search.
And step S6, triggering an alarm system when the support vector machine identifies that the detected object is in a falling state.
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It should be understood by those skilled in the art that the above embodiments do not limit the present invention in any way, and all technical solutions obtained by using equivalent alternatives or equivalent variations fall within the scope of the present invention.
Claims (10)
1. A multi-mode human fall detection method is characterized by comprising the following steps:
acquiring acceleration data of a detection object;
judging the state of the detection object according to the acceleration data;
collecting surface electromyographic signals of the detection object when the detection object is judged to be in a falling state;
extracting electromyographic signal characteristics from the surface electromyographic signal;
identifying the electromyographic signal characteristics through a support vector machine;
and triggering an alarm system when the support vector machine identifies that the detection object is in a falling state.
2. A multi-mode human fall detection method according to claim 1,
and acquiring acceleration data of the detection object through a three-axis acceleration sensor.
3. A multi-mode human fall detection method according to claim 2,
the acceleration data of the detection object collected by the three-axis acceleration sensor comprises acceleration values a of X, Y and Z axesx,ay,az。
4. A multi-mode human fall detection method according to claim 3,
the specific method for judging the state of the detection object according to the acceleration data comprises the following steps:
the supination angle Pitch, the rollover angle Roll, and the left-right rotation angle Yaw are calculated according to the following formulas:
and comparing the supine angle Pitch, the side turning angle Roll and the left and right rotation angle Yaw with corresponding normal values to judge whether the detection object falls down.
5. A multi-mode human fall detection method according to claim 4,
and optimizing the supine angle Pitch, the rollover angle Roll and the left and right rotation angle Yaw through an Aldman filtering algorithm.
6. A multi-mode human fall detection method according to claim 1,
the specific method for extracting the electromyographic signal characteristics from the electromyographic signal comprises the following steps:
preprocessing the surface electromyographic signals;
extracting the electromyographic signal characteristics from the preprocessed surface electromyographic signals through wavelet packet permutation entropy.
7. A multi-mode human fall detection method according to claim 6,
the specific method for preprocessing the surface electromyogram signal comprises the following steps:
and extracting information of a 16Hz-160Hz frequency band from the surface electromyographic signals.
8. A multi-mode human fall detection method according to claim 7,
the specific method for extracting the electromyographic signal characteristics from the preprocessed surface electromyographic signal through the wavelet packet permutation entropy comprises the following steps:
carrying out five-layer wavelet packet decomposition on the surface electromyographic signals;
and calculating the permutation entropy of the surface electromyographic signals after the nine low-frequency subspaces are reconstructed by using a permutation entropy formula to serve as the electromyographic signal characteristics.
9. A multi-mode human fall detection method according to claim 1,
and optimizing a penalty parameter c and a kernel function parameter g of the support vector machine through a Particle Swarm Optimization (PSO).
10. A multi-mode human fall detection method according to claim 9,
the specific method for optimizing the penalty parameter c and the kernel function parameter g of the support vector machine by the particle swarm optimization PSO comprises the following steps:
1) by the pair of learning factors c1，c2Optimizing the weight coefficient, initializing the position and the speed of the particles, and initially setting the initial position of each particle as the initial best position;
2) calculating the fitness of each particle, and evaluating the fitness of the particles by adopting K-fold cross validation;
3) updating the speed and the position of the particle according to a speed and position calculation formula of a PSO algorithm;
the speed and position of the PSO algorithm are calculated as follows:
vij(t+1)＝wvij(t)+c1r1[pij-xij(t)]+c2r2[pgj-xij(t)]，
xij(t+1)＝xij(t)+vij(t+1)，
wherein, c1,c1Is a learning factor, r1,r2Is [0, 1 ]]W is a weight coefficient. p is a radical ofijRepresenting the local optimum position, p, of the particle i in the j-th dimensiongjRepresenting the global optimal position of the particle i in the j dimension; v. ofij(t) represents the flight velocity of the particle i in the j-th dimension in the t-th generation. x is the number ofij(t) represents the current position of the particle i in the j dimension in the t generation;
4) and checking whether a termination condition is met, if so, mapping the group optimal particles into the optimal solution of the penalty parameter c and the kernel function parameter g of the support vector machine, otherwise, turning to the step 2 and continuing a new round of search.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011505755.9A CN112669568A (en) | 2020-12-18 | 2020-12-18 | Multi-mode human body falling detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011505755.9A CN112669568A (en) | 2020-12-18 | 2020-12-18 | Multi-mode human body falling detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112669568A true CN112669568A (en) | 2021-04-16 |
Family
ID=75406530
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011505755.9A Pending CN112669568A (en) | 2020-12-18 | 2020-12-18 | Multi-mode human body falling detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112669568A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103610466A (en) * | 2013-10-17 | 2014-03-05 | 杭州电子科技大学 | EMG fall identification method based on EMD permutation entropy |
| CN104107042A (en) * | 2014-07-10 | 2014-10-22 | 杭州电子科技大学 | Electromyographic signal gait recognition method based on particle swarm optimization and support vector machine |
| CN104127181A (en) * | 2014-07-22 | 2014-11-05 | 杭州电子科技大学 | Electromyographic signal tumble detection method based on WKFDA |
| CN106875631A (en) * | 2017-04-01 | 2017-06-20 | 兰州交通大学 | One kind falls down detection alarm method and system |
| CN107622260A (en) * | 2017-10-26 | 2018-01-23 | 杭州电子科技大学 | Lower limb gait phase identification method based on multi-source bio signal |
| CN107693022A (en) * | 2017-09-27 | 2018-02-16 | 福建工程学院 | A kind of method and device for detecting falling over of human body |
| CN108257352A (en) * | 2017-12-30 | 2018-07-06 | 广州柏颐信息科技有限公司 | A kind of fall detection method for early warning based on intelligent wearable device |
| CN109087482A (en) * | 2018-09-18 | 2018-12-25 | 西安交通大学 | A kind of falling detection device and method |
-
2020
- 2020-12-18 CN CN202011505755.9A patent/CN112669568A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103610466A (en) * | 2013-10-17 | 2014-03-05 | 杭州电子科技大学 | EMG fall identification method based on EMD permutation entropy |
| CN104107042A (en) * | 2014-07-10 | 2014-10-22 | 杭州电子科技大学 | Electromyographic signal gait recognition method based on particle swarm optimization and support vector machine |
| CN104127181A (en) * | 2014-07-22 | 2014-11-05 | 杭州电子科技大学 | Electromyographic signal tumble detection method based on WKFDA |
| CN106875631A (en) * | 2017-04-01 | 2017-06-20 | 兰州交通大学 | One kind falls down detection alarm method and system |
| CN107693022A (en) * | 2017-09-27 | 2018-02-16 | 福建工程学院 | A kind of method and device for detecting falling over of human body |
| CN107622260A (en) * | 2017-10-26 | 2018-01-23 | 杭州电子科技大学 | Lower limb gait phase identification method based on multi-source bio signal |
| CN108257352A (en) * | 2017-12-30 | 2018-07-06 | 广州柏颐信息科技有限公司 | A kind of fall detection method for early warning based on intelligent wearable device |
| CN109087482A (en) * | 2018-09-18 | 2018-12-25 | 西安交通大学 | A kind of falling detection device and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102302370B (en) | Method and device for detecting tumbling | |
| Khan et al. | Accelerometer's position free human activity recognition using a hierarchical recognition model | |
| CN104586398A (en) | Old man falling detecting method and system based on multi-sensor fusion | |
| CN103211599A (en) | Method and device for monitoring tumble | |
| CN112057080B (en) | Freezing gait detection system based on staged feature extraction | |
| CN107609477A (en) | It is a kind of that detection method is fallen down with what Intelligent bracelet was combined based on deep learning | |
| CN108549856A (en) | A kind of human action and road conditions recognition methods | |
| CN108958482B (en) | Similarity action recognition device and method based on convolutional neural network | |
| CN105989694A (en) | Human body falling-down detection method based on three-axis acceleration sensor | |
| CN113100756A (en) | Stacking-based Parkinson tremor detection method | |
| Razum et al. | Optimal threshold selection for threshold-based fall detection algorithms with multiple features | |
| CN107019501B (en) | Remote tumble detection method and system based on genetic algorithm and probabilistic neural network | |
| CN110916672A (en) | Old people daily activity monitoring method based on one-dimensional convolutional neural network | |
| CN112016619A (en) | Fall detection method based on insoles | |
| CN114881079A (en) | Human body movement intention abnormity detection method and system for wearable sensor | |
| CN111626273B (en) | Fall behavior recognition system and method based on atomic action time sequence characteristics | |
| CN107007285B (en) | Fall detection method based on pressure and acceleration transducer | |
| CN114469074A (en) | Fall early warning method, system, equipment and computer storage medium | |
| CN112669568A (en) | Multi-mode human body falling detection method | |
| CN109620241B (en) | Wearable device and motion monitoring method based on same | |
| Liu et al. | Preimpact fall detection for elderly based on fractional domain | |
| CN108648417A (en) | Raspberry Pi old man based on deep learning falls down detecting system | |
| CN114176598A (en) | Multimode-based epilepsy attack early warning method and intelligent bracelet device | |
| WO2022006659A1 (en) | Method and apparatus for detecting fall events | |
| CN108256457B (en) | Heart impact signal deep bottleneck characteristic extraction method based on deep neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210416 |
|
| RJ01 | Rejection of invention patent application after publication |