Title :
sEMG pattern recognition based on GRNN and Adaboost
Author :
Li, Yang ; Tian, Yantao ; Chen, Wanzhong
Author_Institution :
Sch. of Commun. Eng., Jilin Univ., Changchun, China
Abstract :
The artificial neural network is one of the most widely used methods in the multi-channel sEMG pattern recognition. According to the sEMG recognition rate is low, this paper chooses the GRNN neural network, which is effective to solve nonlinear problem, as the classifier; and then by combining it with the Adaboost algorithm to form a strong classifier. In order to guarantee the correct rate of Adaboost algorithm will increase monotonically in the iteration process, a new method of adjusting the weak classifier weights is proposed in this paper, which is to connect the weight adjusting with the recognition ability of current classifier. The experiment shows that comparing with traditional GRNN, GRNN-Adaboost algorithm has better correct rate in classifying six kinds of motion patterns.
Keywords :
electromyography; iterative methods; learning (artificial intelligence); medical signal processing; neural nets; pattern classification; regression analysis; Adaboost algorithm; GRNN neural network; artificial neural network; classifier weights; iteration process; motion pattern classification; multichannel sEMG pattern recognition; nonlinear problem; recognition ability; sEMG recognition rate; Classification algorithms; Educational institutions; Neurons; Pattern recognition; Training; Training data; Wrist; Adaboost; GRNN; neural network; sEMG;
Conference_Titel :
Electronics, Communications and Control (ICECC), 2011 International Conference on
Conference_Location :
Ningbo
Print_ISBN :
978-1-4577-0320-1
DOI :
10.1109/ICECC.2011.6066678
