Using speech acoustics and ML to decode health and affect

This cluster highlights how detailed acoustic analysis of speech—features like pitch, intensity, jitter, shimmer, and cepstral measures—is being used to detect and monitor emotional and clinical states. Work ranges from transformer-based, real-time emotion recognition and CGAN-powered data augmentation for Parkinson’s detection to reviews of speech-derived biomarkers for depression and longitudinal voice assessments. Supporting resources include human vocalization libraries, clinical voice-analysis tools, and studies linking specific acoustic features to emotional vocalizations. Together, these efforts position voice as a rich, non-invasive biomarker for both mental health and neurological disease, powered by modern machine learning.