Multilayer source data fusion with transformer and multilayer perception for emotion recognition

The computing power of computers has been vastly enhanced in recent years due to the swift progress of integrated circuit technology and the chip industry. In the context of decreasing operational costs, the training of artificial intelligence models has become easier. In the case of the breakthrough of NLP large language model-that is, the emergence and wide application of ChatGPT-the progress of computer vision, another area of artificial intelligence research, has attracted more attention. Among them, compared with NLP natural language model, multi-modal emotion recognition related computer vision technology has a unique advantage in human-computer interaction and other fields because of its unique form of data input. In order to further explore the research of multimodal emotion recognition, this paper designs a training model named MSDFTMP (Multilayer source data fusion with transformer and multilayer perception). To attain the goal of emotion categorization of experimental specimens, the experiments demonstrate that this model can significantly enhance the precision of emotion classification.