A HMM-RBFN hybrid classifier for surface electromyography signals classification

Objective:To classify surface electromyography(sEMG) signals by using HMM-RBFN hybrid classifier and to explore the strategy of effectively controlling hand prosthesis. Method: Eight subjects(male 4, female 4) with normal upper limbs were selected in the experiments. Each subject was instructed to perform 7 kinds of fingers movement and each motion was repeated 50 times. The sEMG signals were recorded on 4 forearm muscles. Features of sEMG signals were extracted using wavelet transform and conveyed to HMM classifier and HMM-RBFN hybrid classifier for training. Result: HMM-RBFN hybrid classifier provided better results than that from the single HMM classifier.Conclusion:①The performance of HMM classifier is not so excellent in sEMG signal discrimination.②The HMM-RBFN hybrid classifier combine the advantages of two individual classifiers and offset their disadvantages,hence it achieves higher discrimination, accuracy and stability.