Speech emotion recognition model based on interactive cognitive network

Human expressing emotions through language is a gradually changing process. In order to use the time continuity of speech signals to express specific emotions, this paper builds a GA-GRUS-ICN model based on Interactive cognitive network. Firstly, the GRUS network is used to extract the depth timing features of the input speech features. Then, the self-attention mechanism is introduced to give higher weights to important feature segments. Finally, ICN is used to construct the correlation between emotions to obtain the emotion correlation matrix and the final recognition result. In this paper, the genetic algorithm GA is used to select the hyperparameters. The three basic emotions of “sadness”, “anger” and “happy” in the TYUT2.0, EMO-DB and CASIA emotional voice database are selected as experimental data, the paper designed five experimental schemes to perform two ablation experiments, experimental results show that, The UA of the three models in the three speech database reached 80.83%, 98.61% and 88.13% respectively, indicating that the GA-GRUS-ICN recognition model has strong universality in emotional speech recognition, and the self-attention mechanism is more suitable for the GRUS-ICN model, and can also perform speech emotion recognition well.