Enviar aquest missatge de text: Feature Fusion Based on Main-Auxiliary Network for Speech Emotion Recognition