Towards More Human Machines Speech Emotion Recognition Using Mel-Spectrogram-Based Reconstruction

Speech Emotion Recognition is a field that aims to interpret human emotions from speech signals. By leveraging advanced machine learning and signal processing techniques, this project aims to create a model capable of detecting a range of emotions such as happiness, sadness, anger, fear, surprise, disgust, and neutrality by analyzing speech patterns. Specifically, the project focuses on implementing a self-supervised learning approach through the vector quantized masked autoencoder for speech (VQ-MAE-S), which is fine-tuned for optimal emotion recognition in various contexts.