CN-119884908-B - Lower limb movement identification method based on S-transform energy concentration surface myoelectricity decoding

CN119884908BCN 119884908 BCN119884908 BCN 119884908BCN-119884908-B

Abstract

A lower limb movement identification method based on S transformation energy concentration surface myoelectricity decoding includes the steps of firstly preprocessing an original myoelectric signal, intercepting a preprocessed signal for a useful time period through end point detection, then conducting segmentation S transformation and energy concentration calculation on the signal in the time period, extracting signal characteristics of specified dimensions through segmentation operation, conducting movement pattern classification through SVM, conducting fusion analysis on multi-channel signal characteristics, and conducting lower limb movement identification.

Inventors

XU GUANGHUA
LI BAOYU
LUO DAN
PEI JINJU
XIE JIEREN
YANG ZENGYAO
ZHANG SICONG

Assignees

西安交通大学

Dates

Publication Date: 20260505
Application Date: 20250113

Claims (4)

1. A lower limb movement identification method based on S-transform energy concentration surface myoelectricity decoding is characterized by comprising the following steps: 1) Preprocessing an original electromyographic signal, and intercepting the preprocessed signal for a useful time period through endpoint detection; 2) The signal segmentation S conversion and the energy centralized computation are carried out, and then the segmentation S conversion and the energy centralized computation are carried out on the signals in the time period, and the segmentation operation is adopted to extract the signal characteristics of the appointed dimension; The method comprises the steps of intercepting the data length of an action period as x, setting a sliding window with a window length and a step length, carrying out S-transformation on each piece of data, setting a frequency interval to obtain a time-frequency matrix, carrying out superposition summation on 1/q root values of absolute values of the signals subjected to S-transformation in a time-frequency domain, carrying out q-party operation on summation results, obtaining higher energy concentration degree on the signals with rapid change of frequency components by optimizing the width of a window function in the S-transformation, wherein the energy concentration calculation is shown in a formula (2), decomposing the whole time-frequency concentration degree by adopting an energy concentration measurement method as shown in a formula (3), and improving the time-frequency concentration degree by utilizing the window width at a discrete point; (1) (2) (3) wherein: -signal energy concentration; -time; -frequency; -signal S transformation matrix, -F-discrete frequency range, -CM 1 -time-frequency concentration, Q, a nonlinear parameter, controlling the focusing degree, wherein the larger q is, the stronger the high value region is; -normalizing the time-frequency concentration level, —— In the form of a normalization of (c), -Generalized time-frequency concentration; -low power parameters, emphasizing the global distribution, -High power parameters, emphasizing local concentrations, -Is (are) The normalization form ensures the scale consistency of the time-frequency representation; 3) Constructing a multi-classifier of a support vector machine and analyzing the correlation of multi-channel electromyographic signals; The multi-channel electromyographic signal correlation analysis specifically comprises the steps of searching for the difference of classification effects of single-channel signal fusion and multi-channel signal fusion by analyzing the correlation of electromyographic signals, finding out the optimal lower limb muscle combination, calculating the correlation among multi-channel signal characteristics of rectus femoris, medial femoral muscle, biceps femoris, semitendinous, tibialis anterior and medial gastrocnemius according to different similarity conditions among different tested lower limb muscles, selecting channels with good classification effects to fuse according to correlation analysis results, performing fusion analysis on S-transformation energy concentration characteristics of rectus femoris, biceps femoris and medial gastrocnemius by comparing the classification results of single-channel signals, distributing fusion coefficients by a control variable method, and inputting the characteristics with different weights into a classifier to perform motion classification; 4) And performing motion mode classification through SVM, performing fusion analysis on the multichannel signal characteristics, and performing lower limb motion recognition.
2. The method for identifying the lower limb movement according to claim 1, wherein the signal preprocessing in the step 1) comprises 1.1) 50Hz notch removal of power frequency interference, 1.2) 30Hz zero-phase shift high-pass filtering removal of movement artifacts, and the zero-phase shift filter is adopted in the preprocessing process to preserve the phase information of the original data sequence.
3. The method for identifying the lower limb movement according to claim 1, wherein the detection of the active segment in the step 1) adopts a double threshold detection method based on short-time energy and short-time variance sum, firstly sets a minimum threshold of the initial short-time energy and variance and a low threshold representing fluctuation, and both parameters are dynamically adjusted according to actual conditions; assume that the myoelectric signal of the nth frame is expressed as The frame length is N, and the short-time energy is calculated The variance sum is calculated as shown in formula (4) As shown in the formula (5), (4) (5) In the formula, Is a sample index of the signal, representing sample points in each frame, Is from 0 to An integer of 1, representing the position of each sample point in the frame signal, Representing the mean of the signal.
4. The method for identifying the lower limb movement according to claim 1, wherein the support vector machine multi-classifier construction in the step 3) is characterized in that a multi-classifier is designed, for a given m classes, a classifier is trained for each two classes of the m classes, the total number of the classifiers is m (m-1)/2, and for data needing to be classified, prediction of all the classifiers is needed, and the final class attribute is determined by using a voting mode.

Description

Lower limb movement identification method based on S-transform energy concentration surface myoelectricity decoding Technical Field The invention relates to the technical field of pattern recognition of man-machine interaction, in particular to a lower limb movement recognition method based on S-transform energy concentration surface myoelectricity decoding. Background Along with the continuous development of man-machine interaction technology, the requirements of amputees are changed into intelligent artificial limbs due to the fact that heavy passive artificial limbs are changed into intelligent artificial limbs, the artificial limbs can recognize the movement intention of a human body, the output moment or the joint angle is actively regulated and controlled, the gait of a healthy person is approximate, and the artificial limb is more comfortable to use. The existing lower limb prostheses use mechanical sensing signals (Huang Pingao. Sensing and identifying of movement intention of intelligent lower limb prostheses key technical research [ D ]. University of chinese academy of sciences (advanced technical research institute of shenzhen, academy of chinese sciences), 2020.), but the inherent time lag of mechanical signals causes the control process to lag behind human gait. Part of artificial limbs are controlled by electromyographic signals, but at present, the simple time-frequency domain characteristics （Sánchez-Velasco LE, Arias-Montiel M, Guzmán-Ramírez E, Lugo-González E. A low-cost EMG-controlled anthropomorphic robotic hand for power and precision grasp. Biocybern Biomed Eng 2020;40:221–37.）, of the multi-extraction signals of the electromyographic codes of the lower limbs lose the phase information which is indispensable in the analysis of the movement pattern, so that the identification effect of the movement pattern of the lower limbs of the human body of the lower limbs is poor, and the natural interaction between the intention of a patient and the control of the artificial limbs cannot be realized. Therefore, a method for identifying the movement intention of the human body through decoding the electromyographic signals on the surface of the lower limb needs to be explored, the accuracy of identifying the movement intention of the lower limb is further improved, and the movement control mode of the lower limb artificial limb is enriched. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide a lower limb movement identification method based on S-transformation energy concentration surface myoelectricity decoding, which is used for carrying out segmentation S-transformation on signals so as to preserve phase information of the signals and optimize the calculation process of energy concentration to improve movement classification accuracy, and is used for constructing an SVM multi-classifier, so that the lower limb movement pattern identification accuracy is further improved through S-transformation energy concentration and characteristic fusion thereof. In order to achieve the above object, the present invention adopts the following technical scheme: A lower limb movement recognition method based on S transformation energy concentration surface myoelectricity decoding includes the steps of firstly preprocessing an original myoelectric signal, intercepting a useful time period of the preprocessed signal through end point detection, then carrying out segmentation S transformation and energy concentration calculation on the signal in the time period, extracting signal characteristics of specified dimensions through segmentation operation, classifying movement modes through SVM, carrying out fusion analysis on multichannel signal characteristics, and carrying out lower limb movement recognition. A lower limb movement identification method based on S-transform energy concentration surface myoelectricity decoding comprises the following steps: 1) Signal preprocessing and active segment detection; 2) S conversion and energy concentration calculation of signal segments; 3) Constructing a multi-classifier of a support vector machine and analyzing the correlation of multi-channel electromyographic signals; 4) Support vector machine motion classification based on multichannel signal feature fusion analysis. The signal preprocessing in the step 1) comprises 1.1) 50Hz notch, 1.2) 30Hz zero-phase shift high-pass filtering, and the zero-phase shift filter is adopted in the preprocessing process to keep the phase information of the original data sequence. The method comprises the steps of 1) adopting a double threshold detection method based on short-time energy and short-time variance sum to detect an active segment, firstly setting a minimum threshold of the initial short-time energy and variance sum low threshold representing fluctuation, and dynamically adjusting two parameters according to actual conditions; assume that the myoelectric signal of the nth f