Saifelden M. Ismail
This study introduces a mobile-efficient speech emotion recognition system using DistilHuBERT, achieving significant parameter reduction while maintaining competitive accuracy through cross-corpus validation.
Researchers have developed a new system for recognizing emotions in speech that is efficient enough to run on mobile devices. By using a simplified and compact version of a transformer model, the system reduces the computational load significantly compared to traditional models, without losing much accuracy. This was tested using various datasets to ensure it can generalize well across different types of data. The system shows promise for practical use in mobile applications, balancing the need for small size and high accuracy.
