Multi-modal depression detection method based on hierarchical cross attention mechanism

The invention discloses a multi-modal depression detection method based on a hierarchical cross attention mechanism, and relates to the technical field of artificial intelligence, the method comprises the steps: obtaining an interview video of a user, carrying out the feature extraction of the interview video, obtaining an audio feature and a video feature, inputting the audio feature and the video feature into a trained Transformer model, and obtaining a multi-modal depression detection result; outputting the depression degree of the user; a Transformer model adopts a multi-level cross attention mechanism, data acquisition of the multi-mode depression detection method is combined with a PHQ-8 depression scale with good application reliability, data of a video mode and an audio mode are comprehensively acquired, cross-mode feature information is considered, potential relation between different modes is supplemented, and the detection method has the advantages that the detection accuracy is high, and the detec