应用于假手的肌电信号分类方法研究

Study of Pattern Recognition Methods for Prosthetic Hand Control

高效、准确的肌电信号分类方法是实现高仿生性肌电假手控制的关键技术之一.本文对肌电假手常用的信号分类方法进行了系统研究,阐述了目前常用的七种方法的基本原理和研究现状.重点分析了基于神经网络和支持向量机的肌电信号分类方法,分别从网络类型、所采用的信号特征、识别准确率等方面对分类结果进行了比较.最后,对各种方法的优缺点及适应性进行了分析,对实现高质量的肌电信号识别方法的研究方向进行了探讨.

In order to obtain the purpose of controlling the myoelectric prosthetic hand efficiently and accurately,good signal classification methods must be designed.Classification methods of EMG are studied systematically in this paper.The basic principles and current research status of seven methods which are used frequently are presented,and classification methods of EMG based on neural network and support vector machine are analyzed in particular,and their classification results are compared from the aspects of Network type,signal feature,recognition accuracy and so on.Then,the advantages and disadvantages of each method are discussed as well as their adaptability.Finally,research directions of high quality identification method of EMG are discussed.