Voice emotion recognition method and system based on joint penalty sparse representation dictionary learning
暂无分享,去创建一个
The invention discloses a voice emotion recognition method and system based on the joint penalty sparse representation dictionary learning. The method comprises the steps that 1, feature extraction and processing are carried out on each emotion voice signal in a training sample bank, and a training sample feature matrix A is obtained; 2, the training sample feature matrix A is learnt through a dictionary learning method with the joint penalty of sub-coding and full coding to obtain a dictionary; 3, feature extraction and processing are carried out on the emotion voice signals in the testing sample bank, and each testing sample obtains the corresponding testing sample feature vector y; 4, the testing sample feature vectors y are coded on the dictionary in a sparse mode to obtain a coding coefficient; 5, the coding coefficient is recognized according to the recognition criterion of dictionary learning. The sparse representation dictionary learning method based on the joint penalty of sub-coding and full coding is successfully used for recognizing the emotion voice signals, and the recognition result is more accurate.