Intelligent myoelectric pattern recognition system of 11 hand motions using ant colony optimisation method

Abstract

The selection of optimal coefficients of the feature vector FV is an important step to improve the classification accuracy in myoelectric pattern recognition PR system. In this study, the utilisation of feature selection based on a novel ant colony optimisation ACO approach was investigated to recognise 11 hand motions. The ACO algorithm was employed to choose the best subsets of two extracted features: root mean square RMS and energy of wavelet packet coefficients EWPCs. The optimal selected subsets were utilised as an input vector of radial basis function neural network RBFNN. The highest classification accuracy rate of 94.54% was obtained using the ACO-RBFNN classifier based on selected subsets. The proposed method shows better performance compared with regression tree classifier REGTREE, naive Bayes classifier NavieBayes and K-nearest neighbour K-NN. The average accuracy rate was decreased by 3% when 50% of white Gaussian noise was added to the acquired sEMG signal.