Feature parameters extraction and affective computing of voice message for social media environment

Abstract

Voice message in social media environment includes a large number of conversation natural languages, which increases the difficulty of emotion tagging and affective computing. In order to solve the above difficulties, this paper analyzes the cognitive differences between the semantic and acoustic features of voice message from the perspective of cognitive neuroscience, and presents a voice feature extraction method based on EEG (Electroencephalogram) experiments, and gets the representation of 25 acoustic feature parameter vectors. Meanwhile, we proposed an affective computing method based on PAD (Pleasure-Arousal-Dominance) dimension emotional space according to the above parameters. Experiments show that the method can effectively solve the affective computing problem of voice message. Overall, there are two main contributions of this paper. Firstly, it comprehensively analyzes the emotional cognitive feature of voice message in social media environment from the perspectives of cognitive neural mechanism, voice acoustic feature and text semantics. Secondly, the segmented affective computing method for voice message based on acoustic feature parameters and PAD emotional state model is proposed.