Thesis Project on Speech Emotion Recognition
AudioEmotion Recognition
I engineered and curated a high-quality Bangla speech emotion recognition (SER) dataset to address data scarcity for underrepresented languages. I fine-tuned state-of-the-art transformer models and developed a custom CNN to extract and analyze emotional features from audio data. I conducted comparative analyses across deep learning architectures to validate model performance.
• Collected and labeled Bangla speech audio clips for emotion recognition.
• Fine-tuned pre-trained models (Wav2Vec2, ExHuBERT) on custom and existing datasets.
• Trained convolutional neural networks on Mel spectrograms to identify emotional signals.
• Evaluated and documented model performance across multiple architectures.
I engineered and curated a high-quality Bangla speech emotion recognition (SER) dataset to address data scarcity for underrepresented languages. I fine-tuned state-of-the-art transformer models and developed a custom CNN to extract and analyze emotional features from audio data. I conducted comparative analyses across deep learning architectures to validate model performance. • Collected and labeled Bangla speech audio clips for emotion recognition. • Fine-tuned pre-trained models (Wav2Vec2, ExHuBERT) on custom and existing datasets. • Trained convolutional neural networks on Mel spectrograms to identify emotional signals. • Evaluated and documented model performance across multiple architectures.
2023 - 2023