DFWe: Efficient knowledge distillation of fine-tuned Whisper encoder for speech emotion recognition

Key Points

• The use of knowledge distillation significantly improves the performance of speech emotion recognition models with fine-tuned whisper encoders.
• Accuracy rates improved by 15% in recognizing emotions, demonstrating the effectiveness of the proposed method.
• Analysis involved a neural network architecture, focusing on the efficiency of the fine-tuning process and knowledge transfer.
• Highlights the need for efficient algorithms in speech emotion recognition, potentially enhancing future applications.