Abstract: It is a hard problem that how to find the effective feature from high-dimensional feature in EMU signal emotion recognition. I}his paper used random forest algorithm to comput the contribution in different emotion recognition of the 126 EMU signal, depending on the feature evaluation criteria of random forest algorithm, and then preferentially made up the features that have more contribution to emotion recognition,and used then to emotion recognition. Experiments show it is reasonable.

Key words: Emotion recognition, Random forest, EMG signal, Feature selection