Invited Academic Seminar Series – Sri Krishnan

Abstract: Audio signals such as speech, breathing, cough, and joint sounds provide valuable insights for non-invasive health monitoring and digital health. This talk highlights advances in audio signal analysis spanning six generations of classical signal processing, feature extraction, and AI-driven approaches for real-world healthcare and telemedicine applications.

A particular focus is on the work of the Signal Analysis Research (SAR) Group at Toronto Metropolitan University, which has developed dedicated datasets and models for biomedical audio and audio scene analysis, including speech, cough, respiratory, and knee joint sounds. These datasets and models support systematic benchmarking of feature extraction methods and successive generations of AI models, including those tailored for resource-constrained and edge deployment. The group’s contributions in adaptive signal representations, audio texture analysis, and intelligent sensing and filtering techniques are highlighted as key enablers of robust performance in noisy and variable real-world environments.

The talk will also discuss auditory display techniques, including sonification and audification, as complementary tools for interpreting complex biomedical signals. In addition, multimodal extensions that integrate audio with other sensing modalities will be outlined to illustrate pathways toward more robust, context-aware, and clinically meaningful digital health and telemedicine systems.