Inclusive ASR
Develop inclusive technology for pathological speech
Dysarthric speech faces significant challenges due to data scarcity, which hampers the application of recently developed neural network models.
One promising approach to address this issue is the use of data augmentation techniques for dysarthric speech. I have recently initiated a collaboration with my labmate Chin-Jou Li, my colleagues Dr. Rohan Kumar Das, Prof. Zhengjun Yue, Dr. Paula Andrea Pérez-Toro, my long-time research idols Juan Rafael Orozco Arroyave, Elmar Nöth, and my advisor Prof. David Mortensen. Our first work can be found here: (Li et al., 2025).
References
2025
- InterspeechTowards Inclusive ASR: Investigating Voice Conversion for Dysarthric Speech Recognition in Low-Resource LanguagesIn Interspeech, 2025