This research adapts OpenAI's Whisper model to recognize different types of speech stress patterns across diverse speaker populations, including neurodivergent individuals.

• Specialized recognition of phrasal, lexical, and contrastive stress in speech
• Inclusive dataset featuring 66 native English speakers across genders and neurotypes
• Clinical applications potentially enabling better speech analysis tools for medical assessment
• Advanced ASR capabilities that more closely mirror human speech perception

This breakthrough matters for medical professionals by potentially creating more accessible tools for neurological assessment through speech pattern analysis, opening new avenues for inclusive speech technology.

Fine-Tuning Whisper for Inclusive Prosodic Stress Analysis