Electromyographic signal noise reduction and classification method based on generative adversarial network

Abstract

The invention discloses an electromyographic signal noise reduction and classification method based on a generative adversarial network. The electromyographic signal noise reduction and classificationmethod comprises the steps of preprocessing electromyographic signals and then constructing an electromyographic signal noise reduction generative adversarial network model based on a WGAN; inputtingthe electromyographic signals into an electromyographic signal noise reduction generative adversarial network model for training, minimizing a JS distance between generative distribution and real data distribution, realizing mapping of noise-containing signals and the electromyographic signals, and outputting the electromyographic signals after noise reduction; converting the format of the electromyographic signals after noise reduction into a two-dimensional digital matrix, and extracting electromyographic signal features from the two-dimensional digital matrix by adopting a multi-scale convolution kernel convolution neural network model; important information is selected from two directions of a channel and a space according to an attention mechanism; and finally, the pooling data is tiled, and the electromyographic signals are classified by utilizing a Softmax classifier.

Description

technical field [0001] The invention belongs to the technical field of noise reduction and classification of surface electromyographic signals, and in particular relates to a method for noise reduction and classification of electromyographic signals based on a generative confrontation network. Background technique [0002] Surface electromyography is an electrical signal collected by surface electrodes on the skin of the human body. This electrical signal is the potential difference generated by muscle movement near the muscle fibers. When the human body produces a movement intention, the intention is generated in the brain, encoded in the nerve signal and transmitted to the spinal cord, and then transmitted to the corresponding limb (such as the upper limb) through the nerve pathway after the second encoding. The nerve signal causes the muscle fiber to contract and generate a potential difference. Pull the bone to complete the motion. In this process, the motor intention i...

AI Technical Summary

Problems solved by technology

However, because the EMG signal is relatively weak, it is easily affected by power frequency, ECG, and other noises in life. The EMG signal collected often contains a lot of noise signals, which cannot accurately reflect muscle movements. great inconvenience

Images