Deep Learning Based Emotion Recognition Algorithm for Digital Music Speech

摘要

Verbal communication is the most direct and easiest way for human beings to express their desires and feelings. With the progress of science and technology, the service robot industry has shown rapid development. In this case, voice interaction between people and service robots has become a norm. How to make service robots effectively recognize human voices and emotions when communicating with people is currently a hot issue in the field of human-human communication and service robots. Aiming to solve the current hotspot problem, this paper explores a convolutional neural network (CRN)-based emotion recognition method for digital music speech. The experimental results show that compared with other emotion recognition methods such as CRNN (Convolutional Recurrent Neural Network) and RNN (Recurrent Neural Network), the method proposed in this paper is more effective in music emotion recognition. It can be compared and analyzed with DBC (Deep Bisimulation Control) algorithm, Fisher's method, PCA (Principal Component Analysis) method and Factor Analysis method. The study also proves that the CNN-based algorithm outperforms other algorithms for emotion classification (accuracy rates of 78%, 81%, 65%, 81%, and 65% for “sadness”,”joy”, “lyricism”,”excitement”, and “vent” in speech emotions are higher than those of the other techniques, respectively).