Abstract
The recognition way of linear discrimination and non-linear discrimination is studied in the paper. Classifiers are designed based on linear and non-linear discrimination. Classifying 8 kinds of motion patterns’ surface electromyography acquisitioned from 4 pieces of muscles of the human body forearm was achieved.
Linear classifiers and tree classifier are designed based on Fisher guideline. Subsection linear classifier is designed based on multi-classes arithmetic. Training and testing classifiers by using experiment data, multi-classes recognition strategies are proposed based on two classes arithmetic recognition. Results of classifying different surface electromyography show that the designed classifiers can classify different movements. The linear classifiers and tree classifier’ recognition is higher than non-linear classifiers. The rate of every kinds of movements attain over 80%.
The classifier is designed on the MATLAB 6.5.