Multimodal Emotion Recognition Method Based on Adaptive Selection Strategy

In order to grasp students' emotional state in time, adjust teaching methods in time to stimulate students' learning interest and improve classroom teaching efficiency, this paper proposes an emotion recognition method based on adaptive selection strategy. Firstly, three kinds of emotion signals are acquired and pre-processed and converted to obtain the image signals of three channels. Then, Residual Deep Convolutional Neural Network is used to extract emotion features and preliminary identification of emotion categories for the image signals of multiple channels respectively. Then, adaptive selection strategy is adopted to select effective channels from the preliminary identification results and carry out the fusion of feature parameters. Finally, multi-classifier decision making algorithm is used for final emotion recognition. This method can select the channel data reflecting the true emotion to make the final decision, which makes up the deficiency of single mode and improves the robustness and recognition rate of the recognition method.