Abstract: According to the physiology of myoelectric signals, a two-channel single input multiple output (SIMO) FIR model is proposed to the two-channel upper limb EMG signal in this paper. As the input of the model is unknown and unaccessible, blind signal processing technique is employed to identify the model parameters. After extracting two channels'impulse responses, four types of forearm motions, i.e. hand grasp, hand extension, forearm supination and forearm pronation are able to be classified. Experimental results demonstrate that this method has a better classification result than the classical AR parameter method. This paper shows a promising application of blind signal processing method to the analysis of physiological signals.