Gait emotion recognition method and system based on multi-head fusion

The invention discloses a gait emotion recognition method and system based on multi-head fusion, and relates to the technical field of artificial intelligence, and the method comprises the following steps: obtaining a to-be-predicted video, and carrying out the preprocessing of the to-be-predicted video, and obtaining a fusion feature; inputting the fusion features into an emotion prediction model to output a gait emotion prediction result; according to the gait emotion recognition method, multiple emotion features and data are fused in a crossed mode, the gait emotion recognition accuracy is improved through a multi-head integration method, and a computer vision system can be helped to capture and understand the emotion state of an individual more comprehensively. 本发明公开了一种基于多头融合的步态情绪识别方法及系统，涉及人工智能技术领域，包括如下步骤：获取待预测视频，对待预测视频进行预处理得到融合特征；将融合特征输入到情感预测模型中，以输出步态情绪预测结果；该步态情绪识别方法将多种情感特征与数据交叉融合，通过多头集成的方法，提高了步态情绪识别的准确率，可以更帮助计算机视觉系统更全面地捕捉和理解个体的情感状态。