CN108875563A - A kind of human motion recognition method based on muscle signal - Google Patents
A kind of human motion recognition method based on muscle signal Download PDFInfo
- Publication number
- CN108875563A CN108875563A CN201810399880.2A CN201810399880A CN108875563A CN 108875563 A CN108875563 A CN 108875563A CN 201810399880 A CN201810399880 A CN 201810399880A CN 108875563 A CN108875563 A CN 108875563A
- Authority
- CN
- China
- Prior art keywords
- signal
- neuromuscular junction
- muscle
- human
- sheep
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
Abstract
A kind of human motion recognition method based on muscle signal is claimed in the present invention, obtains the corresponding depth image data of muscle signal based on depth signal stream, identifies the action message of each neuromuscular junction point in three dimensions in human body；Action message based on neuromuscular junction optimizes human muscle's joint signal, filtering does not have an impact on Human bodys' response or influences lesser neuromuscular junction signal or redundant muscular joint signal, human body behavior conditional random field models are constructed, Human bodys' response model is obtained and the subsequent action of human body is predicted based on this model.The present invention selects it using improved flock of sheep optimization algorithm, improved flock of sheep optimization algorithm is relative to unmodified method, improved method is to have used the method initialization population of point set, it avoids algorithm and falls into local optimum, convergence speed of the algorithm is accelerated with the method for order subset and the method for looking around of introducing flock of sheep, while also improving the recognition effect of video human behavior.
Description
Technical field
The invention belongs to mode identification technologies, in particular it relates to which a kind of human body based on muscle signal is dynamic
Make recognition methods.
Background technique
With the rapid development of economy, continuous improvement of people's living standards and computer science, control engineering, rehabilitation doctor
Technology is constantly brought forth new ideas, and disabled person gradually recognizes to make up " incomplete " animation by artificial limb no longer remote.And
The psychological need for how making the response disabled person that artificial limb is more practical, sees, has become the scientific research cause to promote the well-being of mankind.Intelligence is imitative
Thus raw artificial limb comes into being, it can more simulate normal human body posture compared with general artificial limb, to the physiology and the heart of disabled person
Reason has incomparable positive influence.
Although muscle signal also has its disadvantage, more other control signals more being capable of connection freely and control deformed limb and vacation
Limb.So researcher is acquired it and identifies often using muscle signal as the external voltage input of artificial limb.Believed according to muscle
Number acquisition mode can be classified as pin electrode acquisition and surface electrode acquisition.Wherein, although pin electrode to the acquisition of signal more
To be accurate, but due to its by needle electrode to signal acquisition while to body have compared with macrolesion and to picker require compared with
It is more, so researcher generallys use not damaged, the convenient surface electrode of acquisition and is acquired to muscle signal.
In recent years, it was developed to improve muscle signal detection and many relevant technologies of recognition capability, it is main to study
In terms of direction concentrates on following three：More movements are identified using less acquisition channel；It is special to extract more useful signals
Sign；Select more effective classifier identification signal.
However in the prior art, it need to be improved in the information analysis ability for issuing muscle signal, for muscle signal
Data-handling capacity it is also lower, can not accurately and effectively identify movement representated by muscle signal.
Summary of the invention
In order to solve the above technical problems, a kind of human action identification side based on muscle signal is claimed in the present invention
Method.
A kind of human motion recognition method based on muscle signal, it is characterised in that：
Step 1, based on the corresponding depth image data of depth signal stream acquisition muscle signal, in obtained depth image data
Each pixel information include three-dimensional space depth information, the white point in data reprocessed, and then identify
The action message of each neuromuscular junction point in three dimensions in human body；
Step 2, is normalized muscle and dimension-reduction treatment, and the movement coordinate of each neuromuscular junction is subtracted connection muscle
Act coordinate, i.e.,：What is respectively indicated is first neuromuscular junction to N
The action message of a neuromuscular junction；
Step 3, the action message based on neuromuscular junction optimize human muscle's joint signal, and filtering does not produce Human bodys' response
It is raw to influence or influence lesser neuromuscular junction signal or redundant muscular joint signal；
Step 4 clusters as K posture using by Feature Descriptor, the posture after quantization is established model with discrete Markov
And classify；
Step 5 indicates the vision word in bag of words using the action message of neuromuscular junction point, there is shown each word
All it is expressed as the human action for having stronger identification, the word represented further according to these movement examples goes out in dictionary
Existing frequency obtains a histogram about vision, finally as the input of classifier, identification maneuver；
Step 6, constructs human body behavior conditional random field models, and training sample obtains Human bodys' response model and based on the mould
Type predicts the subsequent action of human body.
The present invention selects it using improved flock of sheep optimization algorithm, and improved flock of sheep optimization algorithm is not relative to
Improved method, improved method are to have used the method initialization population of point set, avoid algorithm and fall into local optimum, with having
The method of sequence subset and the method for looking around for introducing flock of sheep accelerate convergence speed of the algorithm, while also improving video human behavior
Recognition effect.
Detailed description of the invention
It is included to provide the attached drawing further recognized to published subject, this specification will be incorporated into and constitute this and said
A part of bright book.Attached drawing also illustrates the realization of published subject, and disclosed for explaining together with detailed description
The realization principle of theme.It is not attempt to show to be more than the knot needed to the basic comprehension of published subject and its a variety of practice modes
Structure details.
A kind of work flow diagram for human motion recognition method based on muscle signal that attached drawing 1 is protected for the present invention.
Specific embodiment
Advantages of the present invention, feature and reach the method for the purpose will be bright by attached drawing and subsequent detailed description
Really.
A kind of work flow diagram for human motion recognition method based on muscle signal that attached drawing 1 is protected for the present invention.
A kind of human motion recognition method based on muscle signal, it is characterised in that：
Step 1, based on the corresponding depth image data of depth signal stream acquisition muscle signal, in obtained depth image data
Each pixel information include three-dimensional space depth information, the white point in data reprocessed, and then identify
The action message of each neuromuscular junction point in three dimensions in human body；
Step 2, is normalized muscle and dimension-reduction treatment, and the movement coordinate of each neuromuscular junction is subtracted connection muscle
Act coordinate, i.e.,：What is respectively indicated is action message of first neuromuscular junction to the N neuromuscular junction；
Step 3, the action message based on neuromuscular junction optimize human muscle's joint signal, and filtering does not produce Human bodys' response
It is raw to influence or influence lesser neuromuscular junction signal or redundant muscular joint signal；
Step 4 clusters as K posture using by Feature Descriptor, the posture after quantization is established model with discrete Markov
And classify；
Step 5 indicates the vision word in bag of words using the action message of neuromuscular junction point, there is shown each word
All it is expressed as the human action for having stronger identification, the word represented further according to these movement examples goes out in dictionary
Existing frequency obtains a histogram about vision, finally as the input of classifier, identification maneuver；
Step 6, constructs human body behavior conditional random field models, and training sample obtains Human bodys' response model and based on the mould
Type predicts the subsequent action of human body.
Specifically, muscle is normalized in the step 2 and dimension-reduction treatment, the movement of each neuromuscular junction is sat
Mark subtracts the movement coordinate of connection muscle, i.e.,：What is respectively indicated is
One neuromuscular junction to the N neuromuscular junction action message, including：
Selecting a human body 3D joint coordinates is master pattern；
B) it keeps each sample limb segment direction vector constant, each vector is zoomed into master pattern length
It is in the action definition of t frame, limbs i：, wherein i{ 1 ..., N }, N
What is indicated is the number of artis, directly using the joint action information with time change as the Expressive Features of behavior.
Specifically, the step 3, the action message based on neuromuscular junction optimize human muscle's joint signal, filter to people
Body Activity recognition does not have an impact or influences lesser neuromuscular junction signal or redundant muscular joint signal further includes：
Screening and filtering optimization processing is carried out to neuromuscular junction signal based on flock of sheep algorithm,
Step 3.1 initializes the group containing N sheep, the initial population being evenly distributed with good point set.Setting maximum changes
Generation number, wherein defining flock of sheep in the minimum identification space of D dimension space search food is [0,0 ... 0], maximum identification space
For [1,1 ... 1]；
Step 3.2, the fitness value for assessing initial population, are set as 0 for the algebra of population；
Step 3.3 judges whether the algebra of population is greater than maximum the number of iterations, if it is greater, then stopping calculating, output phase is answered
Maximum adaptation angle value (accuracy of identification), otherwise turn to the 4th step；
Step 3.4 judges that can the number of iterations of population divide exactly G（What G was indicated is the algebra that a flock of sheep keep relationship, experiment
According to experiment experience be taken as 10）If divided exactly, the 3.5th step is turned to, otherwise turns to the 3.6th step；
Step 3.5 sorts according to obtained fitness value, and establishes a hierarchy, obtains an order subset.By one
Flock of sheep are divided into several groups and determine the relationship of lamb and sheep mother；
Step 3.6, the movement more new formula that ram is obtained according to formula 4.1 are substituted according to OS method with formula 4.6
The action message of action message in formula 4.1, calculating fitness value, lamb and ewe is constant.If obtained fitness value
Bigger than instead not preceding fitness value, then substituted 4.1, otherwise do not substitute.Wherein, the value of b is taken as 0.25.According to formula
4.3 and 4.5 respectively obtain the movement more new formula of ewe and lamb, and the movement with ram updates processing unanimously；Wherein, b
Value be taken as 0.2 and 0.1 (by the experience value of experiment) respectively.Turn to the 3.7th step；
Step 3.7 updates the current optimal movement of individual in flock of sheep and the optimal movement of overall situation individual of flock of sheep, and the number of iterations adds 1, and
Turn to the 3rd step；
The current optimal movement of individual and the optimal movement of overall situation individual of flock of sheep are to influence the action recognition in step 3.8, flock of sheep
Maximum neuromuscular junction signal and whole neuromuscular junction ensemble.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the present invention.
Claims (3)
1. a kind of human motion recognition method based on muscle signal, it is characterised in that：
Step 1, based on the corresponding depth image data of depth signal stream acquisition muscle signal, in obtained depth image data
Each pixel information include three-dimensional space depth information, the white point in data reprocessed, and then identify
The action message of each neuromuscular junction point in three dimensions in human body；
Step 2, is normalized muscle and dimension-reduction treatment, and the movement coordinate of each neuromuscular junction is subtracted connection muscle
Coordinate is acted, i.e.,：What is respectively indicated is first neuromuscular junction to N
The action message of neuromuscular junction；
Step 3, the action message based on neuromuscular junction optimize human muscle's joint signal, and filtering does not produce Human bodys' response
It is raw to influence or influence lesser neuromuscular junction signal or redundant muscular joint signal；
Step 4 clusters as K posture using by Feature Descriptor, the posture after quantization is established model with discrete Markov
And classify；
Step 5 indicates the vision word in bag of words using the action message of neuromuscular junction point, there is shown each word
All it is expressed as the human action for having stronger identification, the word represented further according to these movement examples goes out in dictionary
Existing frequency obtains a histogram about vision, finally as the input of classifier, identification maneuver；
Step 6, constructs human body behavior conditional random field models, and training sample obtains Human bodys' response model and based on the mould
Type predicts the subsequent action of human body.
2. a kind of human motion recognition method based on muscle signal as described in claim 1, it is characterised in that：The step
Muscle is normalized in two and dimension-reduction treatment, the movement coordinate of each neuromuscular junction is subtracted to the movement coordinate of connection muscle, i.e.,：What is respectively indicated is first neuromuscular junction to the N muscle
The action message of connector, including：
Selecting a human body 3D joint coordinates is master pattern；
B) it keeps each sample limb segment direction vector constant, each vector is zoomed into master pattern length；
It is in the action definition of t frame, limbs i：, wherein i{ 1 ..., N }, N table
What is shown is the number of artis, directly using the joint action information with time change as the Expressive Features of behavior.
3. a kind of human motion recognition method based on muscle signal as described in claim 1, it is characterised in that：The step
Three, action message based on neuromuscular junction optimizes human muscle's joint signal, filtering Human bodys' response is not had an impact or
It influences lesser neuromuscular junction signal or redundant muscular joint signal further includes：
Screening and filtering optimization processing is carried out to neuromuscular junction signal based on flock of sheep algorithm,
Step 3.1 initializes the group containing N sheep, the initial population being evenly distributed with good point set
Set maximum number of iterations, wherein defining flock of sheep in the minimum identification space of D dimension space search food is [0,0 ...
0], maximum identification space is [1,1 ... 1]；
Step 3.2, the fitness value for assessing initial population, are set as 0 for the algebra of population；
Step 3.3 judges whether the algebra of population is greater than maximum the number of iterations, if it is greater, then stopping calculating, output phase is answered
Maximum adaptation angle value (accuracy of identification), otherwise turn to the 4th step；
Step 3.4 judges that can the number of iterations of population divide exactly G（What G was indicated is the algebra that a flock of sheep keep relationship, experiment
According to experiment experience be taken as 10）If divided exactly, the 3.5th step is turned to, otherwise turns to the 3.6th step；
Step 3.5 sorts according to obtained fitness value, and establishes a hierarchy, obtains an order subset
One flock of sheep is divided into several groups and determines the relationship of lamb and sheep mother；
Step 3.6, the movement more new formula that ram is obtained according to formula 4.1 are substituted according to OS method with formula 4.6
The action message of action message in formula 4.1, calculating fitness value, lamb and ewe is constant
If obtained fitness value is bigger than instead not preceding fitness value, otherwise substituted 4.1 does not substitute
Wherein, the value of b is taken as 0.25
The movement more new formula that ewe and lamb are respectively obtained according to formula 4.3 and 4.5, the movement with ram update processing one
It causes；Wherein, the value of b is taken as 0.2 and 0.1 (by the experience value of experiment) respectively
Turn to the 3.7th step；
Step 3.7 updates the current optimal movement of individual in flock of sheep and the optimal movement of overall situation individual of flock of sheep, and the number of iterations adds 1, and
Turn to the 3rd step；
The current optimal movement of individual and the optimal movement of overall situation individual of flock of sheep are to influence the action recognition in step 3.8, flock of sheep
Maximum neuromuscular junction signal and whole neuromuscular junction ensemble.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810399880.2A CN108875563A (en) | 2018-04-28 | 2018-04-28 | A kind of human motion recognition method based on muscle signal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810399880.2A CN108875563A (en) | 2018-04-28 | 2018-04-28 | A kind of human motion recognition method based on muscle signal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108875563A true CN108875563A (en) | 2018-11-23 |
Family
ID=64326925
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810399880.2A Pending CN108875563A (en) | 2018-04-28 | 2018-04-28 | A kind of human motion recognition method based on muscle signal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108875563A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670513A (en) * | 2018-11-27 | 2019-04-23 | 西安交通大学 | A kind of piston attitude detecting method based on bag of words and support vector machines |
| CN111832528A (en) * | 2020-07-24 | 2020-10-27 | 北京深醒科技有限公司 | Detection and abnormity analysis method for behavior analysis |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105095862A (en) * | 2015-07-10 | 2015-11-25 | 南开大学 | Human gesture recognizing method based on depth convolution condition random field |
| CN103810496B (en) * | 2014-01-09 | 2017-01-25 | 江南大学 | 3D (three-dimensional) Gaussian space human behavior identifying method based on image depth information |
| CN106650619A (en) * | 2016-11-15 | 2017-05-10 | 上海师范大学 | Human action recognition method |
| CN107901040A (en) * | 2017-12-03 | 2018-04-13 | 陈思佳 | Robot myoelectric control system based on ROS |
-
2018
- 2018-04-28 CN CN201810399880.2A patent/CN108875563A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103810496B (en) * | 2014-01-09 | 2017-01-25 | 江南大学 | 3D (three-dimensional) Gaussian space human behavior identifying method based on image depth information |
| CN105095862A (en) * | 2015-07-10 | 2015-11-25 | 南开大学 | Human gesture recognizing method based on depth convolution condition random field |
| CN106650619A (en) * | 2016-11-15 | 2017-05-10 | 上海师范大学 | Human action recognition method |
| CN107901040A (en) * | 2017-12-03 | 2018-04-13 | 陈思佳 | Robot myoelectric control system based on ROS |
Non-Patent Citations (2)
| Title |
|---|
| 唐凤: "人体行为识别中骨骼特征信息的鸡群优化选择算法", 《数码设计》 * |
| 唐凤: "基于Kinect骨骼的视频人体行为识别", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670513A (en) * | 2018-11-27 | 2019-04-23 | 西安交通大学 | A kind of piston attitude detecting method based on bag of words and support vector machines |
| CN111832528A (en) * | 2020-07-24 | 2020-10-27 | 北京深醒科技有限公司 | Detection and abnormity analysis method for behavior analysis |
| CN111832528B (en) * | 2020-07-24 | 2024-02-09 | 北京深醒科技有限公司 | Detection and anomaly analysis method for behavior analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kamel et al. | Deep convolutional neural networks for human action recognition using depth maps and postures | |
| CN106650687B (en) | Posture correction method based on depth information and skeleton information | |
| CN110705390A (en) | Body posture recognition method and device based on LSTM and storage medium | |
| Liu et al. | Computational model based on neural network of visual cortex for human action recognition | |
| CN108875586B (en) | Functional limb rehabilitation training detection method based on depth image and skeleton data multi-feature fusion | |
| Yang et al. | Towards zero re-training for long-term hand gesture recognition via ultrasound sensing | |
| WO2021243561A1 (en) | Behaviour identification apparatus and method | |
| CN107132915A (en) | A kind of brain-machine interface method based on dynamic brain function network connection | |
| CN108875563A (en) | A kind of human motion recognition method based on muscle signal | |
| Zhang et al. | Research on sEMG-based gesture recognition by dual-view deep learning | |
| CN115346272A (en) | Real-time tumble detection method based on depth image sequence | |
| Chao et al. | CZU-MHAD: a multimodal dataset for human action recognition utilizing a depth camera and 10 wearable inertial sensors | |
| Zhou et al. | Intelligent recognition of medical motion image combining convolutional neural network with Internet of Things | |
| Liu et al. | Trampoline motion decomposition method based on deep learning image recognition | |
| Meng et al. | Listen to your face: Inferring facial action units from audio channel | |
| Chen et al. | Cnn-lstm model for recognizing video-recorded actions performed in a traditional chinese exercise | |
| Su | Implementation and rehabilitation application of sports medical deep learning model driven by big data | |
| Shi et al. | Accurate and efficient LIF-Nets for 3D detection and recognition | |
| Xu et al. | A Novel Event-Driven Spiking Convolutional Neural Network for Electromyography Pattern Recognition | |
| Xu et al. | Isolated Word Sign Language Recognition Based on Improved SKResNet-TCN Network | |
| Pan et al. | Combining pose and trajectory for skeleton based action recognition using two-stream RNN | |
| Ji et al. | Analysis of the Impact of the Development Level of Aerobics Movement on the Public Health of the Whole Population Based on Artificial Intelligence Technology | |
| Zhang et al. | Automatic construction and extraction of sports moment feature variables using artificial intelligence | |
| Li et al. | 2D multi-person pose estimation combined with face detection | |
| Li et al. | Tai chi action quality assessment and visual analysis with a consumer rgb-d camera |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20201125 Address after: No.7082, Heping District e-commerce building, No.16 Lhasa Road, nanyingmen street, Heping District, Tianjin Applicant after: Mango thinking (Tianjin) Intelligent Technology Co., Ltd Address before: 300457 Tianjin City Binhai New Area Tianjin Development Zone, Tianjin Development Zone Xin Huan West Road, No. 8 Taida Service Outsourcing Industrial Park 2 floor (Tianjin coastal Service Outsourcing Industry Co., Ltd. trusteeship No. 2889th) Applicant before: SHANGGU TECHNOLOGY (TIANJIN) Co.,Ltd. Applicant before: Han Junlai Applicant before: Hu Yana Applicant before: Xie Jin |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181123 |