Skip to Main Content
In this paper,we use both wavelet transform and BP neural network to identify SEMG from human upper arm. In the experiments,we decompose a single action into different parts to realize the multi-level identification of a single action. We use two electrodes to extract SEMG signal from the upper arm biceps,triceps firstly,then analyze this signal using wavelet transform and extract eight values forming the feature vector,finally put this feature vectors into BP neural network to complete pattern recognition. The results of the experiments using the method introduced in this paper show that the average recognition rate of arm internal rotating, external rotating, arm stretching, 1/3 bending, 1/2 bending and full bending is over 90%.