Speech emotion recognition method and system

The invention relates to a speech emotion recognition method and system, and the method comprises the following steps: S1, carrying out the parallel channel attention weighting of an input speech feature map, and obtaining a weighted feature map; s2, performing feature extraction on the weighted feature map through small-size residual connection and feature convolution to obtain a deep fusion emotion feature map; performing shallow feature extraction on the weighted feature map through large-scale residual connection to obtain a shallow emotion feature map; and S3, fusing the deep-fusion emotion feature map with the shallow-layer emotion feature map to obtain a weighted-fusion emotion feature map. The emotion features in the speech feature map are fully extracted, loss of global features is reduced, extraction of fused emotion features is effectively achieved, meanwhile, the model complexity is low, and the problems of slow training, low efficiency and the like caused by too complex networks are solved. 本发明涉及