Traditional speech emotion analysis techniques are influenced by factors such as environment and speech speed. Given this, this study assumes the combination of grid particle swarm optimization algorithm and support vector regression, combined with a pleasure-arousal-dominance emotion dimension model, to improve the accuracy of speech emotion analysis. This paper first constructs a speech emotion model that combines the pleasure-arousal-dominance dimensions of particle swarm optimization and support vector regression. Then, a comparative analysis is conducted on the performance of the sentiment analysis model in this dimension and its applicability in the direction of broadcasting hosting. In the experiment of standard deviation and average value of pleasure-arousal-dominance dimension, the scoring average curve of the dimension of the research algorithm was almost a straight line, and the components of pleasure-arousal-dominance were normally distributed. The minimum standard deviation scores of pleasures, dominance, and arousal were 0.40, 0.56, and 0.82. The results demonstrate that the model exhibits high reliability in all dimensions of pleasure-arousal-dominance prediction, thereby substantiating its potential as a pivotal tool in the analysis of speech emotions exhibited by broadcast hosts.
Read full abstract