CN111476158A - Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM - Google Patents
Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM Download PDFInfo
- Publication number
- CN111476158A CN111476158A CN202010263776.8A CN202010263776A CN111476158A CN 111476158 A CN111476158 A CN 111476158A CN 202010263776 A CN202010263776 A CN 202010263776A CN 111476158 A CN111476158 A CN 111476158A
- Authority
- CN
- China
- Prior art keywords
- physiological signal
- svm
- frequency
- somatosensory
- pso
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- 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
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM, Step 1: collecting an original sample of a human physiological signal; step 2: extracting physiological signal characteristics; step 3: analyzing and extracting key features of the principal components; step 4: establishing a multi-channel physiological signal somatosensory gesture SVM recognition model; step 5: PSO optimizes and trains a multi-channel physiological signal somatosensory gesture PCA and SVM recognition model. The invention provides a multi-channel physiological signal somatosensory gesture recognition method based on a PSO-PCA-SVM, which is high in recognition accuracy, good in real-time performance and good in recognition robustness.
Description
Technical Field
The invention belongs to the field of gesture recognition, and particularly relates to a multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM.
Background
With the development of economy and the improvement of technology level, somatosensory Interaction (gesture Interaction) is changing the understanding of people on the design of traditional products as a novel Interaction mode rich in behavior ability, and exploring a novel behavior mode. The somatosensory interaction is an interaction mode which directly utilizes body motion, sound, eyeball rotation and other modes to interact with peripheral devices or environments, and the life quality of people is greatly improved. Only when the interactive system recognizes the somatosensory gesture of people, the functions which people want to achieve can be executed, so the somatosensory gesture recognition technology is an important component in somatosensory interaction, but is also one of research difficulties.
The traditional method uses a somatosensory recognition technology based on computer vision, but the method cannot completely meet the requirement of somatosensory gesture recognition due to the reasons of strong privacy invasiveness, limited observation range, easiness in being influenced by various factors such as illumination conditions and shielding. With the development of sensor technology, somatosensory gesture recognition based on a sensor is gradually common, for example, a kinect is utilized to acquire a depth image and skeleton information to recognize the somatosensory gesture, great progress is obtained in gesture recognition, but the kinect still can be influenced by a shielding environment. In recent years, somatosensory gesture recognition technology based on physiological signals has come to emerge. The physiological signals are slightly influenced by environmental factors, the real-time performance is strong, and after the characteristics are extracted, the somatosensory gestures can be well represented.
Disclosure of Invention
To solve the above existing problems. The invention provides a multi-channel physiological signal somatosensory gesture recognition method based on a PSO-PCA-SVM, which is high in recognition accuracy, good in real-time performance and good in recognition robustness. To achieve this object:
the invention provides a multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM, which comprises the following specific steps:
step 1: collecting an original sample of a human physiological signal;
selecting physiological signals of two channels of surface electromyographic signals and electroencephalographic signals as samples for somatosensory gesture recognition, selecting a plurality of volunteers with healthy body conditions as experimental research objects, wherein each experimenter is provided with wearable intelligent interactive equipment, the equipment is provided with a surface electromyographic sensor sEMG and an electroencephalographic sensor EEG, the experimenter respectively carries out corresponding action gestures for a plurality of times, and the sensors respectively collect the surface electromyographic signals and the electroencephalographic signals of each gesture and mark the gestures according to the types of the gestures;
step 2: extracting physiological signal characteristics;
the physiological signals are used as data sources for body sensing gesture recognition, and representative features need to be extracted from the physiological signals to serve as important indexes in body sensing gesture classification.
Measuring the characteristics of physiological signal data by selecting time domain characteristics and frequency domain characteristics of the physiological signal, wherein the number of the time domain characteristics is 7, and the expression of the mean value is
Where N is the sample length of the signal sequence,
is normalized physiological signal value
Wherein xminIs the minimum value of the physiological signal, xmaxIs the maximum value of the physiological signal, xkA kth value representing a physiological signal;
the standard deviation is expressed as
The first-order difference can represent the speed of signal change and the change trend and the pole value existing in the signal, and the expression is
Wherein, tkIs a sampling time node of a physiological signal;
the second-order difference can detect the inflection point position in the signal, and the expression is
Before extracting the frequency domain characteristics of the physiological signals, processing the signals by adopting Fourier transform FFT (fast Fourier transform) to convert time domain signals into frequency domain signals, extracting the characteristics of the obtained frequency domain signals, wherein the number of the extracted frequency domain characteristics is 3, the median frequency represents the frequency of a frequency spectrum divided into two regions with equal amplitude, and the expression is
pjIs the power spectrum of the muscle electricity signal at the frequency j, M is the length of the whole frequency band, the mean frequency represents the ratio of the sum of the products of the power spectrum and the frequency to the sum of the spectral intensity, and the expression is
Wherein f isjRepresenting the frequency spectrum over a frequency band, the frequency ratio being the ratio of the low frequency component to the high frequency component of the physiological signal, expressed as
Wherein U L C and LL C are upper truncation frequencies and lower truncation frequencies of a low frequency band, UHC and L HC are upper truncation frequencies and lower truncation frequencies of a high frequency band, and 20 time domain and frequency domain characteristics of a surface electromyogram signal and an electroencephalogram signal of each somatosensory gesture sample are extracted together;
step 3: analyzing and extracting key features of the principal components;
because a certain correlation exists among a plurality of physiological signal characteristics, the Principal Component Analysis (PCA) is adopted to reduce the dimension of the multidimensional characteristics to remove cross redundant information so as to extract key characteristics and construct a physiological signal sample matrix X ═ X1,X2,X3,...,X20]Wherein each column XiRepresenting a sample characteristic dimension, calculating a training average
And the difference di＝XiPsi, constructing a covariance matrix
Wherein A ═ d1,d2,...,d20]Solving for AA using SVD theorem of singular value decompositionTAnd arranging λ in monotonically decreasing order1≥λ2≥…≥λpThe corresponding feature vectors are respectively mu1,μ2,...,μp(p is less than or equal to n), and selecting the first p maximum eigenvectors to form a linear transformation matrix W ═ mu1,μ2,...,μp]Projecting the original input feature vector into a p-dimensional subspace, PCi＝WTdiWherein PCiFor the first i principal component arrays, the array matrix PC is respectively divided into1,PC2,...,PC20After normalization processing, the samples are sequentially used as SVM model input samples;
step 4: establishing a multi-channel physiological signal somatosensory gesture SVM recognition model;
the SVM algorithm is based on a statistical learning theory and a structure risk minimum principle, has stronger high-dimensional sample processing capacity, adopts an optimal classification plane method to classify different types, converts a gesture recognition problem into a plurality of binary problems through a multi-channel physiological signal somatosensory gesture SVM recognition model, constructs 5 SVM two classifiers, determines a sample with the gesture class of K as a positive sample in the K-th classification, combines the rest other types of gesture samples as a negative sample, thus obtaining a large two classifier, realizes the purpose of recognizing 4 types of somatosensory gesture samples, finds an optimal hyperplane meeting the data classification requirement on the classification basis of each somatosensory gesture SVM two classifier, ensures that the hyperplane has the maximum distance with two types of sample points under the condition of ensuring the classification precision, and can be represented as w.phi (x) + b as 0, w is a weight vector, b is an offset, and based on a statistical theory, the support vector machine model determines a classification function through minimization of the following target numbers:
s.t.yi[wTxi+b]≥1-ξi,(ξi≥0,i＝1,...,l)
wherein C is a penalty parameter, the penalty degree of the wrong sample can be controlled, ξiIntroducing Lagrange multiplication operators for relaxation factors, solving the above formula, and establishing an objective function for finding an optimal hyperplane:
partial derivatives are respectively calculated for lambda and b, and are made to be equal to zero, linear classification after nonlinear transformation is realized by adopting a proper inner product function, inner product operation among training samples is realized, and lambda and b can be solved. The final classification function of the SVM is established as
Wherein λ is Lagrange multiplier, K (X)iX) is kernel function of SVM model, RBF kernel function is selected, and expression K (X)i,x)＝exp(-g||Xi-x||2) Wherein g is a nuclear parameter;
step 5: PSO optimizes and trains a multi-channel physiological signal somatosensory gesture PCA and SVM recognition model;
the Particle Swarm Optimization (PSO) algorithm is adopted to obtain the optimal principal component parameter matrix PCiThe values of a kernel function parameter g and a penalty parameter C;
firstly, a real number coding mode is adopted, the size of a selected population is 30, and the iteration number is 100. Determining a position boundary [ X ]min,Xmax]And velocity boundary [ V ]min,Vmax]Randomly initializing the position and speed of each particle in the population, training by combining a sample set, wherein the optimized objects are recognition rate and recognition time, and calculating a fitness value
WhereinL is the number of training samples, Time is the Time required for identifying the training sample set, ACC is the accuracy rate of identifying the training sample set,timein order to identify the weight coefficients for the time,accto identify the weight coefficient of the accuracy, and at the same time, to obtain the individual optimal position PpAnd the group optimal position PgAnd their corresponding individual extreme values and group extreme values, updating the particle velocity Vi,k+1＝wVi,k+c1r1(Pp,k-Xi,k)+c2r2(Pg,k-Xi,k) In which V isi,k+1Is the velocity of the ith particle in the kth iteration, c1、c2Are acceleration factors, r1、r2Is a random number between 0 and 1, if the calculated value exceeds the speed range Vmin,Vmax]The boundary value is used instead to update the particle position Xi,k+1＝Xi,k+Vi,k+1Wherein X isi,k+1For the position of the ith particle in the kth iteration, if the calculated value exceeds the velocity range [ X ]min,Xmax]Replacing the optimal particle position with a boundary value, recalculating the fitness value, updating the example position corresponding to the extreme value, ending PSO optimization if the error meets the precision requirement, and obtaining the optimized optimal particle position as the optimized principal component parameter matrix PCiThe kernel function parameter g and the punishment parameter C are combined with a sample set for training to obtain an optimal PSO optimization SVM multi-channel physiological signal somatosensory gesture recognition model, the whole model algorithm is based on an MAT L AB platform to write a program, the SVM model is called through a L IBSVM tool, and sample data training and testing are uploaded on the platform;
step 6: testing based on an optimal multi-channel physiological signal somatosensory gesture recognition model;
after an optimal PSO-PCA-SVM multi-channel physiological signal somatosensory gesture recognition model is established, electroencephalogram signals and surface electromyogram signals of a human body are collected in real time, and after feature extraction, gesture recognition results are obtained through the PSO-PCA-SVM multi-channel physiological signal somatosensory gesture recognition model.
As a further improvement of the present invention, the step1 action gestures include a fist, a spread finger, a wave in and a wave out gesture.
As a further improvement of the invention, the time domain characteristics in step2 are 7, which are respectively the maximum value Max, the minimum value Min, the Median, the Mean, the standard deviation Std, the first-order difference 1Diff and the second-order difference 2 Diff.
As a further improvement of the present invention, the 3 Frequency domain features extracted in step2 are median Frequency, MDF, Mean Frequency, MNF, and Frequency Ratio, FR, respectively.
The invention provides a multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM, which has the following beneficial effects:
1) the method uses the multi-channel physiological signals as the characteristics of somatosensory gesture recognition, has better real-time performance, is less influenced by the environment, and has good recognition robustness.
2) The method extracts the frequency domain characteristics and the time domain characteristics of the multi-channel physiological signals, enriches the characteristic quantity and lays a solid foundation for establishing a model for representing the somatosensory gestures.
3) The invention adopts the PCA principal component analysis method to reduce the dimension of the features, retains the key features, reduces the relevance of the features of the input subsequent model, and greatly reduces the complexity of the SVM recognition model.
4) The method utilizes the global search optimal characteristic of the particle swarm algorithm to improve and optimize the support vector machine model and the PCA model, aims at maximizing the recognition accuracy and minimizing the recognition time, obtains the optimal model parameters, has better generalization capability, and saves a large amount of manual parameter selection work.
5) Compared with the traditional method, the multi-channel physiological signal somatosensory gesture recognition method based on the PSO-PCA-SVM has better recognition accuracy and can meet the actual application requirements.
Drawings
FIG. 1 is a flow chart of a multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM of the invention;
FIG. 2 is a training algorithm of a multichannel physiological signal somatosensory gesture recognition optimal model based on PSO optimized PCA-SVM of the invention;
FIG. 3 illustrates the support vector machine model of the present invention.
Detailed Description
The invention is described in further detail below with reference to the following detailed description and accompanying drawings:
the invention provides a multi-channel physiological signal somatosensory gesture recognition method based on a PSO-PCA-SVM, which is high in recognition accuracy, good in real-time performance and good in recognition robustness.
As an embodiment of the invention, a flow chart of a multi-channel physiological signal somatosensory gesture recognition method based on a PSO-PCA-SVM is shown in fig. 1, a training algorithm of a multi-channel physiological signal somatosensory gesture recognition optimal model based on a PSO-optimized PCA-SVM is shown in fig. 2, a support vector machine model is shown in fig. 3, and specific steps are as follows.
Step 1: collecting human physiological signal original sample
The method selects physiological signals of two channels of surface electromyographic signals and electroencephalographic signals to make a somatosensory gesture recognition sample. A plurality of volunteers with healthy physical conditions are selected as experimental research objects, each experimenter is worn with wearable intelligent interaction equipment, and the equipment comprises a surface electromyogram (sEMG) sensor and an electroencephalogram (EEG). The experimenter makes hand gestures such as fist (fist), hand (spread finger), inward swing (wave in) and outward swing (wave out) for multiple times, and the sensor collects surface myoelectric signals and electroencephalogram signals of the hand gestures at each time and marks the hand gestures according to the hand gesture types. To ensure that the physiological signals collected under each gesture can be accurately labeled, only the middle 10s of the signals of about 12s per gesture are labeled as steady-state signals. It eliminates the transition state between two gestures and also avoids the corresponding time to follow the next gesture.
Step 2: physiological signal feature extraction
The physiological signals are used as data sources for body sensing gesture recognition, and representative features need to be extracted from the physiological signals to serve as important indexes in body sensing gesture classification.
In the invention, the time domain characteristics and the frequency domain characteristics of the physiological signals are selected to measure the characteristics of the physiological signal data, wherein the number of the time domain characteristics is 7, and the time domain characteristics are respectively a maximum value (Max), a minimum value (Min), a Median value (Median), a Mean value (Mean), a standard deviation (Std), a first-order difference (1Diff) and a second-order difference (2 Diff).
The expression of the mean value is
Where N is the sample length of the signal sequence,
is normalized physiological signal value
Wherein xminIs the minimum value of the physiological signal, xmaxIs the maximum value of the physiological signal, xkRepresenting the kth value of the physiological signal.
The standard deviation is expressed as
The first-order difference can represent the speed of signal change and the change trend and the pole value existing in the signal, and the expression is
Wherein, tkIs a sampling time node of the physiological signal.
The second-order difference can detect the inflection point position in the signal, and the expression is
Before extracting the frequency domain characteristics of the physiological signals, processing the signals by adopting Fourier transform (FFT) to convert the time domain signals into frequency domain signals, and obtaining the frequency domain signalsAnd extracting features from the frequency domain signal. There are 3 Frequency domain features extracted, which are the Median Frequency (MDF), Mean Frequency (MNF), and Frequency Ratio (FR). The median frequency represents the frequency at which the spectrum is divided into two equal-amplitude regions, and is expressed as
pjTo be the myoelectric signal power spectrum at frequency j, M is the length of the entire band. The mean frequency represents the ratio of the sum of the products of the power spectrum and the frequency to the sum of the intensities of the spectrum, and is expressed as
Wherein f isjRepresenting the spectrum over a frequency band. The frequency ratio is the ratio of the low frequency component and the high frequency component of the physiological signal, and the expression is
Wherein U L C and LL C are the upper and lower cut-off frequencies of the low band, and UHC and L HC are the upper and lower cut-off frequencies of the high band.
And extracting 20 time domain and frequency domain characteristics of the surface electromyogram signal and the electroencephalogram signal of each somatosensory gesture sample.
Step 3: principal component analysis extracting key characteristics
Because a certain correlation exists among a plurality of physiological signal characteristics, a Principal Component Analysis (PCA) method is adopted to reduce the dimension of the multidimensional characteristics to remove cross redundant information so as to extract key characteristics. The principal component analysis method is a multivariate statistical method for investigating the correlation among a plurality of variables, and researches how to disclose the internal structure among the plurality of variables through a few principal components, namely, the few principal components are derived from the original variables, so that the information of the original variables is kept as much as possible and are not mutually correlated, and the dimension is reduced.
Constructing a physiological signal sample matrix X ═ X1,X2,X3,...,X20]Wherein each column XiRepresenting a sample feature dimensionAnd (4) degree. Calculating a training average
And the difference di＝XiPsi, constructing a covariance matrix
Wherein A ═ d1,d2,...,d20]. Solving for AA using singular value decomposition theorem (SVD)TAnd arranging λ in monotonically decreasing order1≥λ2≥…≥λpThe corresponding feature vectors are respectively mu1,μ2,...,μp(p.ltoreq.n). Selecting the first p maximum eigenvectors to form a linear transformation matrix W ═ mu1,μ2,...,μp]Projecting the original input feature vector into a p-dimensional subspace, PCi＝WTdiWherein PCiThe first i main component groups. Respectively combining the array matrixes PC1,PC2,…,PC20After normalization, the process proceeds
Samples are input for the SVM model.
Step 4: establishing multichannel physiological signal somatosensory gesture SVM recognition model
The SVM algorithm is based on a statistical learning theory and a structure risk minimum principle, has strong high-dimensional sample processing capacity, and realizes classification of different classes by adopting an optimal classification plane method. The invention provides a multichannel physiological signal somatosensory gesture SVM recognition model, which is characterized in that firstly, a gesture recognition problem is converted into a plurality of binary classification problems, 5 SVM two classifiers are constructed, the Kth classification takes a sample with a gesture class of K as a positive sample, and the rest other classes of gesture samples are combined to be a negative sample, so that a large two classifier is obtained, and the purpose of recognizing 4 classes of somatosensory gesture samples is realized. The classification basis of each somatosensory gesture SVM two-classifier is to find an optimal hyperplane meeting the data classification requirement, so that the distance between the hyperplane and two types of sample points is the largest under the condition that the classification precision of the hyperplane is ensured. The classification hyperplane can be expressed as w · Φ (x) + b being 0, w being the weight vector and b being the offset. Based on statistical theory, the support vector machine model determines the classification function by minimizing the number of targets:
s.t.yi[wTxi+b]≥1-ξi,(ξi≥0,i＝1,...,l)
wherein C is a penalty parameter, the penalty degree of the wrong sample can be controlled, ξiIs a relaxation factor. Introducing a Lagrange multiplication operator, solving the formula, and establishing an objective function for searching the optimal hyperplane:
partial derivatives are respectively calculated for lambda and b, and are made to be equal to zero, linear classification after nonlinear transformation is realized by adopting a proper inner product function, inner product operation among training samples is realized, and lambda and b can be solved. The final classification function of the SVM is established as
Wherein λ is Lagrange multiplier, K (X)iX) is a kernel function of the SVM model, the selection of the invention is an RBF kernel function, and an expression K (X)i,x)＝exp(-g||Xi-x||2) Wherein g is a nuclear parameter.
Step 5: PSO (particle swarm optimization) optimization training multi-channel physiological signal somatosensory gesture PCA (principal component analysis) and SVM (support vector machine) recognition model
In the PCA model and the SVM model respectively established at Step3 and 4, a principal component parameter matrix PCiThe selection and the numerical values of the kernel function parameter g and the penalty parameter C have great influence on the accuracy of the model, and the method adopts a Particle Swarm Optimization (PSO) algorithm to obtain an optimal principal component parameter matrix PCiAnd the values of kernel function parameter g and penalty parameter C. The PSO algorithm is an evolutionary computing technology (evolutionary computing) and is derived from the behavior research of bird population predation, and the basic idea is that the behavior research is carried out through the interpersonal behaviors of individuals in the populationCollaboration and information sharing to find the optimal solution. Firstly, a real number coding mode is adopted, the size of a selected population is 30, and the iteration number is 100. Determining a position boundary [ X ]min,Xmax]And velocity boundary [ V ]min,Vmax]The position and velocity of each particle in the population is randomly initialized. Training is carried out by combining a sample set, the optimal objects of the method are the recognition rate and the recognition time, and the fitness value is calculated
Wherein l is the number of training samples, Time is the Time required by the identification of the training sample set, ACC is the identification accuracy of the training sample set,timein order to identify the weight coefficients for the time,accto identify the weighting factors for the accuracy. At the same time, the individual optimum position P is obtainedpAnd the group optimal position PgAnd their corresponding individual extrema and group extrema. Updating the particle velocity Vi,k+1＝wVi,k+c1r1(Pp,k-Xi,k)+c2r2(Pg,k-Xi,k) In which V isi,k+1Is the velocity of the ith particle in the kth iteration, c1、c2Are acceleration factors, r1、r2Is a random number between 0 and 1, if the calculated value exceeds the speed range Vmin,Vmax]The boundary value is substituted. Updating the particle position Xi,k+1＝Xi,k+Vi,k+1Wherein X isi,k+1For the position of the ith particle in the kth iteration, if the calculated value exceeds the velocity range [ X ]min,Xmax]The boundary value is substituted. After the fitness value is recalculated, the example position corresponding to the extreme value is updated, if the error meets the precision requirement, PSO optimization is ended, and the optimized optimal particle position is obtained and used as the optimized principal component parameter matrix PCiThe whole model algorithm is based on an MAT L AB platform to compile a program, the SVM model is called through a L IBSVM tool, and sample data is loaded on the platform to train and test.
Step 6: somatosensory gesture recognition model test based on optimal multi-channel physiological signals
After an optimal PSO-PCA-SVM multi-channel physiological signal somatosensory gesture recognition model is established, electroencephalogram signals and surface electromyogram signals of a human body are collected in real time, and after feature extraction, gesture recognition results are obtained through the PSO-PCA-SVM multi-channel physiological signal somatosensory gesture recognition model.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made according to the technical spirit of the present invention are within the scope of the present invention as claimed.
Claims (4)
1. A multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM is characterized by comprising the following specific steps:
step 1: collecting an original sample of a human physiological signal;
selecting physiological signals of two channels of surface electromyographic signals and electroencephalographic signals as samples for somatosensory gesture recognition, selecting a plurality of volunteers with healthy body conditions as experimental research objects, wherein each experimenter is provided with wearable intelligent interactive equipment, the equipment is provided with a surface electromyographic sensor sEMG and an electroencephalographic sensor EEG, the experimenter respectively carries out corresponding action gestures for a plurality of times, and the sensors respectively collect the surface electromyographic signals and the electroencephalographic signals of each gesture and mark the gestures according to the types of the gestures;
step 2: extracting physiological signal characteristics;
the physiological signals are used as data sources for somatosensory gesture recognition, and representative features need to be extracted from the physiological signals to serve as important indexes for somatosensory gesture classification;
measuring the characteristics of physiological signal data by selecting time domain characteristics and frequency domain characteristics of the physiological signal, wherein the number of the time domain characteristics is 7, and the expression of the mean value is
Where N is the sample length of the signal sequence,
is normalized physiological signal value
Wherein xminIs the minimum value of the physiological signal, xmaxIs the maximum value of the physiological signal, xkA kth value representing a physiological signal;
the standard deviation is expressed as
The first-order difference can represent the speed of signal change and the change trend and the pole value existing in the signal, and the expression is
Wherein, tkIs a sampling time node of a physiological signal;
the second-order difference can detect the inflection point position in the signal, and the expression is
Before extracting the frequency domain characteristics of the physiological signals, processing the signals by adopting Fourier transform FFT (fast Fourier transform) to convert time domain signals into frequency domain signals, extracting the characteristics of the obtained frequency domain signals, wherein the number of the extracted frequency domain characteristics is 3, the median frequency represents the frequency of a frequency spectrum divided into two regions with equal amplitude, and the expression is
pjIs the power spectrum of the muscle electric signal at the frequency j, M is the length of the whole frequency band, and the mean frequency represents the power spectrum and the frequencyThe ratio of the sum of the products to the sum of the spectral intensities, expressed as
Wherein f isjRepresenting the frequency spectrum over a frequency band, the frequency ratio being the ratio of the low frequency component to the high frequency component of the physiological signal, expressed as
Wherein U L C and LL C are upper truncation frequencies and lower truncation frequencies of a low frequency band, UHC and L HC are upper truncation frequencies and lower truncation frequencies of a high frequency band, and 20 time domain and frequency domain characteristics of a surface electromyogram signal and an electroencephalogram signal of each somatosensory gesture sample are extracted together;
step 3: analyzing and extracting key features of the principal components;
because a certain correlation exists among a plurality of physiological signal characteristics, the Principal Component Analysis (PCA) is adopted to reduce the dimension of the multidimensional characteristics to remove cross redundant information so as to extract key characteristics and construct a physiological signal sample matrix X ═ X1,X2,X3,...,X20]Wherein each column XiRepresenting a sample characteristic dimension, calculating a training average
And the difference di＝XiPsi, constructing a covariance matrix
Wherein A ═ d1,d2,...,d20]Solving for AA using SVD theorem of singular value decompositionTAnd arranging λ in monotonically decreasing order1≥λ2≥…≥λpThe corresponding feature vectors are respectively mu1,μ2,...,μp(p is less than or equal to n), and selecting the first p maximum eigenvectors to form a linear transformation matrix W ═ mu1,μ2,...,μp]Projecting the original input feature vector into a p-dimensional subspace, PCi＝WTdiWherein PCiFor the first i principal component arrays, the array matrix PC is respectively divided into1,PC2,...,PC20After normalization processing, the samples are sequentially used as SVM model input samples;
step 4: establishing a multi-channel physiological signal somatosensory gesture SVM recognition model;
the SVM algorithm is based on a statistical learning theory and a structure risk minimum principle, has stronger high-dimensional sample processing capacity, adopts an optimal classification plane method to classify different types, converts a gesture recognition problem into a plurality of binary problems through a multi-channel physiological signal somatosensory gesture SVM recognition model, constructs 5 SVM two classifiers, determines a sample with the gesture class of K as a positive sample in the K-th classification, combines the rest other types of gesture samples as a negative sample, thus obtaining a large two classifier, realizes the purpose of recognizing 4 types of somatosensory gesture samples, finds an optimal hyperplane meeting the data classification requirement on the classification basis of each somatosensory gesture SVM two classifier, ensures that the hyperplane has the maximum distance with two types of sample points under the condition of ensuring the classification precision, and can be represented as w.phi (x) + b as 0, w is a weight vector, b is an offset, and based on a statistical theory, the support vector machine model determines a classification function through minimization of the following target numbers:
s.t.yi[wTxi+b]≥1-ξi,(ξi≥0,i＝1,...,l)
wherein C is a penalty parameter, the penalty degree of the wrong sample can be controlled, ξiIntroducing Lagrange multiplication operators for relaxation factors, solving the above formula, and establishing an objective function for finding an optimal hyperplane:
partial derivatives of λ, b are calculated separately and made equal to zero, as appropriateThe inner product function realizes linear classification after nonlinear transformation, realizes inner product operation among training samples, can solve lambda and b, and the established final classification function of the SVM is
Wherein λ is Lagrange multiplier, K (X)iX) is kernel function of SVM model, RBF kernel function is selected, and expression K (X)i,x)＝exp(-g||Xi-x||2) Wherein g is a nuclear parameter;
step 5: PSO optimizes and trains a multi-channel physiological signal somatosensory gesture PCA and SVM recognition model;
the Particle Swarm Optimization (PSO) algorithm is adopted to obtain the optimal principal component parameter matrix PCiThe values of a kernel function parameter g and a penalty parameter C;
firstly, a real number coding mode is adopted, the size of a population is selected to be 30, the iteration number is 100, and a position boundary [ X ] is determinedmin,Xmax]And velocity boundary [ V ]min,Vmax]Randomly initializing the position and speed of each particle in the population, training by combining a sample set, wherein the optimized objects are recognition rate and recognition time, and calculating a fitness value
Wherein l is the number of training samples, Time is the Time required by the identification of the training sample set, ACC is the identification accuracy of the training sample set,timein order to identify the weight coefficients for the time,accto identify the weight coefficient of the accuracy, and at the same time, to obtain the individual optimal position PpAnd the group optimal position PgAnd their corresponding individual extreme values and group extreme values, updating the particle velocity Vi,k+1＝wVi,k+c1r1(Pp,k-Xi,k)+c2r2(Pg,k-Xi,k) In which V isi,k+1Is the velocity of the ith particle in the kth iteration, c1、c2Are acceleration factors, r1、r2Is a random number between 0 and 1, if the calculated value exceeds the speed range Vmin,Vmax]The boundary value is used instead to update the particle position Xi,k+1＝Xi,k+Vi,k+1Wherein X isi,k+1For the position of the ith particle in the kth iteration, if the calculated value exceeds the velocity range [ X ]min,Xmax]Replacing the optimal particle position with a boundary value, recalculating the fitness value, updating the example position corresponding to the extreme value, ending PSO optimization if the error meets the precision requirement, and obtaining the optimized optimal particle position as the optimized principal component parameter matrix PCiThe kernel function parameter g and the punishment parameter C are combined with a sample set for training to obtain an optimal PSO optimization SVM multi-channel physiological signal somatosensory gesture recognition model, the whole model algorithm is based on an MAT L AB platform to write a program, the SVM model is called through a L IBSVM tool, and sample data training and testing are uploaded on the platform;
step 6: testing based on an optimal multi-channel physiological signal somatosensory gesture recognition model;
after an optimal PSO-PCA-SVM multi-channel physiological signal somatosensory gesture recognition model is established, electroencephalogram signals and surface electromyogram signals of a human body are collected in real time, and after feature extraction, gesture recognition results are obtained through the PSO-PCA-SVM multi-channel physiological signal somatosensory gesture recognition model.
2. The multi-channel physiological signal somatosensory gesture recognition method based on the PSO-PCA-SVM, according to claim 1, is characterized in that: the step1 action gestures include a fist, a hand spread finger, a wave in and a wave out gesture.
3. The multi-channel physiological signal somatosensory gesture recognition method based on the PSO-PCA-SVM, according to claim 1, is characterized in that: the time domain characteristics in the step2 are 7, namely a maximum value Max, a minimum value Min, a Median, a Mean value Mean, a standard deviation Std, a first-order difference 1Diff and a second-order difference 2 Diff.
4. The multi-channel physiological signal somatosensory gesture recognition method based on the PSO-PCA-SVM, according to claim 1, is characterized in that: the 3 Frequency domain characteristics extracted in the step2 are the Median Frequency, media Frequency, MDF, Mean Frequency, MNF and Frequency Ratio, Frequency, and FR, respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010263776.8A CN111476158B (en) | 2020-04-07 | 2020-04-07 | Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010263776.8A CN111476158B (en) | 2020-04-07 | 2020-04-07 | Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111476158A true CN111476158A (en) | 2020-07-31 |
| CN111476158B CN111476158B (en) | 2020-12-04 |
Family
ID=71749914
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010263776.8A Active CN111476158B (en) | 2020-04-07 | 2020-04-07 | Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111476158B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112641449A (en) * | 2020-12-18 | 2021-04-13 | 浙江大学 | EEG signal-based rapid evaluation method for cranial nerve functional state detection |
| CN113127533A (en) * | 2021-03-31 | 2021-07-16 | 四川省气象服务中心(四川省专业气象台 四川省气象影视中心) | Influence factor analysis method of meteorological traffic system based on combined multivariate correlation |
| CN113887397A (en) * | 2021-09-29 | 2022-01-04 | 中山大学中山眼科中心 | Classification method and classification system of electrophysiological signals based on ocean predator algorithm |
| CN114077298A (en) * | 2020-08-07 | 2022-02-22 | 北京大学 | Non-contact gesture recognition method and system, computer equipment and storage medium |
| CN115715680A (en) * | 2022-12-01 | 2023-02-28 | 杭州市第七人民医院 | Anxiety discrimination method and device based on connective tissue potential |
| CN116893314A (en) * | 2023-09-04 | 2023-10-17 | 国网浙江省电力有限公司余姚市供电公司 | Non-invasive power load monitoring method, device, equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014001058A1 (en) * | 2012-06-25 | 2014-01-03 | Softkinetic Software | Improvements in or relating to three dimensional close interactions |
| US20160206206A1 (en) * | 2015-01-19 | 2016-07-21 | Samsung Electronics Company, Ltd. | Optical Detection and Analysis of Bone |
| 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 |
| CN108309328A (en) * | 2018-01-31 | 2018-07-24 | 南京邮电大学 | A kind of Emotion identification method based on adaptive fuzzy support vector machines |
| CN108614991A (en) * | 2018-03-06 | 2018-10-02 | 上海数迹智能科技有限公司 | A kind of depth image gesture identification method based on Hu not bending moments |
| CN109033954A (en) * | 2018-06-15 | 2018-12-18 | 西安科技大学 | A kind of aerial hand-written discrimination system and method based on machine vision |
| CN109271840A (en) * | 2018-07-25 | 2019-01-25 | 西安电子科技大学 | A kind of video gesture classification method |
| CN110908515A (en) * | 2019-11-27 | 2020-03-24 | 北京航空航天大学 | Gesture recognition method and device based on wrist muscle pressure |
-
2020
- 2020-04-07 CN CN202010263776.8A patent/CN111476158B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014001058A1 (en) * | 2012-06-25 | 2014-01-03 | Softkinetic Software | Improvements in or relating to three dimensional close interactions |
| US20160206206A1 (en) * | 2015-01-19 | 2016-07-21 | Samsung Electronics Company, Ltd. | Optical Detection and Analysis of Bone |
| 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 |
| CN108309328A (en) * | 2018-01-31 | 2018-07-24 | 南京邮电大学 | A kind of Emotion identification method based on adaptive fuzzy support vector machines |
| CN108614991A (en) * | 2018-03-06 | 2018-10-02 | 上海数迹智能科技有限公司 | A kind of depth image gesture identification method based on Hu not bending moments |
| CN109033954A (en) * | 2018-06-15 | 2018-12-18 | 西安科技大学 | A kind of aerial hand-written discrimination system and method based on machine vision |
| CN109271840A (en) * | 2018-07-25 | 2019-01-25 | 西安电子科技大学 | A kind of video gesture classification method |
| CN110908515A (en) * | 2019-11-27 | 2020-03-24 | 北京航空航天大学 | Gesture recognition method and device based on wrist muscle pressure |
Non-Patent Citations (4)
| Title |
|---|
| DIANA C等: "A Study of Movement Classification of the Lower", 《ELECTRONICS》 * |
| 徐斌: "基于脑电与肌电信号融合的多自由度手部动作识别研究", 《万方——中国学位论文全文数据库》 * |
| 胡命嘉等: "基于PSO-SVM的手势识别方法研究", 《长春理工大学学报（自然科学版）》 * |
| 胡杜鹃: "基于表面肌电信号的手指手势分类方法研究", 《中国优秀硕士学位论文全文数据库——信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114077298A (en) * | 2020-08-07 | 2022-02-22 | 北京大学 | Non-contact gesture recognition method and system, computer equipment and storage medium |
| CN114077298B (en) * | 2020-08-07 | 2023-10-10 | 北京大学 | Contactless gesture recognition method, system, computer equipment and storage medium |
| CN112641449A (en) * | 2020-12-18 | 2021-04-13 | 浙江大学 | EEG signal-based rapid evaluation method for cranial nerve functional state detection |
| CN113127533A (en) * | 2021-03-31 | 2021-07-16 | 四川省气象服务中心(四川省专业气象台 四川省气象影视中心) | Influence factor analysis method of meteorological traffic system based on combined multivariate correlation |
| CN113887397A (en) * | 2021-09-29 | 2022-01-04 | 中山大学中山眼科中心 | Classification method and classification system of electrophysiological signals based on ocean predator algorithm |
| CN115715680A (en) * | 2022-12-01 | 2023-02-28 | 杭州市第七人民医院 | Anxiety discrimination method and device based on connective tissue potential |
| CN116893314A (en) * | 2023-09-04 | 2023-10-17 | 国网浙江省电力有限公司余姚市供电公司 | Non-invasive power load monitoring method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111476158B (en) | 2020-12-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111476158B (en) | Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM | |
| CN106371610B (en) | Electroencephalogram signal-based driving fatigue detection method | |
| Chen et al. | Surface electromyography feature extraction via convolutional neural network | |
| Sikder et al. | Human activity recognition using multichannel convolutional neural network | |
| CN113158793B (en) | Multi-class motor imagery electroencephalogram signal identification method based on multi-feature fusion | |
| CN110472649B (en) | Electroencephalogram emotion classification method and system based on multi-scale analysis and integrated tree model | |
| CN112244873A (en) | Electroencephalogram time-space feature learning and emotion classification method based on hybrid neural network | |
| CN114533086B (en) | Motor imagery brain electrolysis code method based on airspace characteristic time-frequency transformation | |
| CN113011239B (en) | Motor imagery classification method based on optimal narrow-band feature fusion | |
| CN109512390B (en) | Sleep staging method and wearable device based on EEG time domain multi-dimensional features and M-WSVM | |
| CN102930284A (en) | Surface electromyogram signal pattern recognition method based on empirical mode decomposition and fractal | |
| CN111797674B (en) | MI electroencephalogram signal identification method based on feature fusion and particle swarm optimization algorithm | |
| Mousa et al. | A novel brain computer interface based on principle component analysis | |
| CN109009098B (en) | Electroencephalogram signal feature identification method under motor imagery state | |
| CN111310656A (en) | Single motor imagery electroencephalogram signal identification method based on multi-linear principal component analysis | |
| Jinliang et al. | EEG emotion recognition based on granger causality and capsnet neural network | |
| CN109685071A (en) | Brain electricity classification method based on the study of common space pattern feature width | |
| CN107045624B (en) | Electroencephalogram signal preprocessing and classifying method based on maximum weighted cluster | |
| CN113553896B (en) | Electroencephalogram emotion recognition method based on multi-feature depth forest | |
| Kumari et al. | Effect of reduced dimensionality on deep learning for human activity recognition | |
| CN111265214B (en) | Electroencephalogram signal analysis method based on data structured decomposition | |
| CN109800651B (en) | Multiclass electroencephalogram classification method based on double-rule active overrun learning machine | |
| Zhang et al. | Wavelet transform and singular value decomposition of EEG signal for pattern recognition of complicated hand activities | |
| Almanza-Conejo et al. | Emotion recognition using time-frequency distribution and GLCM features from EEG signals | |
| Sikder et al. | Heterogeneous hand guise classification based on surface electromyographic signals using multichannel convolutional 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |