My focus is on the development of technology for the automated discovery, measurement, representation and learning of the information encoded in the voice signal for optimal voice intelligence.
I began working in computer speech recognition and general audio processing in 1997. Until 2014, my work spanned language-agnostic speech processing, automated discovery and learning of information from speech, and minimally supervised voice processing designed to be robust in high-noise and complex acoustic environments.
In December 2014, I began building up the science of profiling humans from their voice — the concurrent deduction of myriad physical, physiological, medical, psychological, sociological, behavioral and environmental parameters from the voice signal. Because profiling focuses on the voice signal rather than semantic or pragmatic content, it is agnostic to language.
Current work includes designing AI systems for genetic and biomarker discovery from voice, and for exploring other aspects of the human physical state and psyche through the portal of voice. I also work on core designs for general AI systems capable of universal speech and audio processing, and their extension to embodied AI systems.
Since 2019, recognising that such systems will eventually require much more than classical compute, I have also been working on aspects of quantum computing applied to machine learning.
For a full list of current research, students, and publications, see the MLSP group page, the CVIS page, and her CMU SCS profile.