Fusion-ConvBERT: Parallel Convolution and BERT Fusion for Speech Emotion Recognition.

Area of Science:

• Artificial Intelligence
• Speech Processing
• Machine Learning

Background:

• Traditional speech emotion recognition relies on handcrafted features and manual labels.
• Deep learning offers automatic feature extraction but requires substantial training data.
• Existing methods face challenges due to limited data for training deep networks.

Purpose of the Study:

• To develop a novel deep learning approach for speech emotion recognition that maximizes information exploitation from speech signals.
• To address the challenge of limited data in training deep networks for emotion recognition.

Main Methods:

• Proposed Fusion-ConvBERT, a parallel fusion model combining Bidirectional Encoder Representations from Transformers (BERT) and Convolutional Neural Networks (CNNs).
• Utilized EMO-DB and Interactive Emotional Dyadic Motion Capture Database for model evaluation.

Main Results:

• The Fusion-ConvBERT model demonstrated superior performance compared to state-of-the-art techniques.
• The model achieved high accuracy in speech emotion recognition across various test configurations.

Conclusions:

• Fusion-ConvBERT effectively enhances speech emotion recognition by leveraging deep learning architectures.
• The proposed method offers a promising solution for more natural human-computer interactions, even with limited data.