Portfolio project
Speech Emotion Recognition (Wav2Vec 2.0)
Developed and evaluated a Speech Emotion Recognition system using Wav2Vec 2.0 on the RAVDESS dataset. Compared Vanilla Fine-Tuning, TAPT, and Pseudo-TAPT strategies, reaching 81.60% accuracy and 81.27% weighted F1 with Vanilla Fine-Tuning.
- core technologies
- 5documented project stack
- project status
- Builtportfolio implementation
- delivery scope
- End-to-endfrom implementation to evaluation
01Problem
Developed and evaluated a Speech Emotion Recognition system using Wav2Vec 2.0 on the RAVDESS dataset. Compared Vanilla Fine-Tuning, TAPT, and Pseudo-TAPT strategies, reaching 81.60% accuracy and 81.27% weighted F1 with Vanilla Fine-Tuning.
02Approach
- Designed the solution using Python, PyTorch, Wav2Vec 2.0.
- Implemented the core workflow and evaluated the result against the project objective.
- Documented the work as part of the broader AI and data portfolio.
03Contribution
- Contributed to technical implementation and project delivery.
- Worked across data preparation, modeling or application development as required by the project.
04Outcome
- Developed and evaluated a Speech Emotion Recognition system using Wav2Vec 2.0 on the RAVDESS dataset. Compared Vanilla Fine-Tuning, TAPT, and Pseudo-TAPT strategies, reaching 81.60% accuracy and 81.27% weighted F1 with Vanilla Fine-Tuning.
05Limitations
- Additional technical documentation will be added as the project archive is expanded.