Multi-modal emotion recognition method and device

The invention relates to the technical field of emotion recognition, in particular to a multi-mode emotion recognition method and device, and the method comprises the steps: carrying out the pre-segmentation processing of long-sequence audio and video information, inputting audio and video feature codes, and extracting audio and video segment-level feature sequences; connecting the audio and video segment-level feature sequences and then mapping the audio and video segment-level feature sequences through a full connection layer to obtain a segment-level emotion similarity feature sequence; using each segment-level emotion similarity feature sequence query element and each audio and video segment-level feature sequence as a key element and a value element, and outputting an audio and video segment-level emotion weighted feature sequence through a multi-head attention mechanism; respectively calculating an audio and video weighted center vector and a center vector of emotion similarity information by utilizing