Surface electromyography (SEMG)-based human hand interior action identification method

Abstract

The invention provides a surface electromyography (SEMG)-based human hand interior action identification method. By combing common actions of human hand operation, electromyographic signals of all theactions are obtained through SEMG, the original electromyographic signals are preprocessed through an empirical mode decomposition (EMD) algorithm, and the electromyographic signals subjected to noise reduction processing are subjected to feature extraction through a maximum Lyapunov exponent (MLE) method, and finally obtained nonlinear MLE features are classified through a random forest (RF) algorithm. According to a nonlinear time series analysis method for SEMG signal processing and classifying through EMD, MLE and RF, the experiment result of data analysis and comparison shows that through the SEMG-based human hand interior action identification method, the ten kinds of different human hand interior actions can be identified effectively, and the accuracy rate reaches up to 91.67%.

Description

technical field [0001] The invention belongs to the technical field of surface electromyography signal gesture recognition, and in particular relates to a SEMG-based method for human hand motion recognition. Background technique [0002] With the development of sensor technology and signal processing technology, surface electromyography (SEMG), as a novel sensing technology, has been widely used in the fields of multifunctional prosthetic hand control, clinical medicine, and human-computer interaction. SEMG is to collect the bioelectrical signals of superficial muscle and nerve trunk activities on the skin surface through electrodes, and record, filter, amplify, transmit and feedback them, so as to realize the evaluation and simulation of muscle function. Since the SEMG signals of human hands are weak and susceptible to interference (noise, electromagnetic interference, etc.) during object manipulation, it is difficult to measure them. How to effectively collect signals, ext...

AI Technical Summary

Problems solved by technology

Experimental results show that these classifiers can achieve high action classification accuracy, but more complex structures may increase the time and space complexity of training and testing

In addition, most of them are aimed at the classification of actions caused by single muscle contraction in the absence of an operator, while the multi-muscle contraction combined with the action recognition of objects in the hand, the uncertainty of SEMG signals such as muscle fatigue, different experimenters, etc. have not been considered in the Inside

Images