DrBipasha Kashyap

Research focus:

My research asks how much we can reliably learn about health from the human voice. Speech carries emotional, linguistic and pathological information through a single acoustic channel. My group develops signal processing and representation-learning methods that separate these sources of variation, so that clinical claims made by speech AI rest on measurable evidence. Recent work with Professor Björn Schuller and Professor Pubudu Pathirana quantified dimensional independence in speech across six corpora using mutual information, presented at Interspeech 2026, the flagship conference of the International Speech Communication Association (ISCA) and a CORE A-ranked venue.

Foundational work in ataxic speech:

This programme builds on my foundational work in objective speech assessment for cerebellar ataxia. Since 2017 I have developed automated, clinically interpretable measures of ataxic dysarthria, from syllabic timing deficits and vocal tract biomarkers to magnitude- and phase-based cepstral features. This work led to a machine-learning scoring system that predicts the risk and severity of ataxic speech across speech tasks (IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2023) and a mixed-effects framework that models how speech deficits progress over time (IEEE Access, 2021), which future trials of disease-modifying therapies will need. My most cited paper, in Scientific Reports (2020), combined speech with sensor-based motor measures for a comprehensive objective assessment of the condition. My speech timing and cepstral analysis papers in Biomedical Signal Processing and Control and Annals of Biomedical Engineering laid the groundwork for this programme. This work is cited by the Ataxia Center at Massachusetts General Hospital and Harvard Medical School in its own deep-learning and digital biomarker studies of ataxic speech, published in IEEE Access (2024) and The Cerebellum (2023). I have since extended this work to dysphonia, using quantum machine-learning approaches for small speech datasets (IEEE EMBC 2025).

Clinical translation:

This work has a direct clinical path. Speech-processing methods from the NSBE Research Lab are embedded in the BioKin AIMs medical-grade assessment devices, used across 26 clinics on four continents within the UNIFAI clinical consortium. Current projects extend this to depression and other neurological conditions, using controlled-access clinical corpora, and to wearable and acoustic sensing for non-invasive cardiovascular monitoring.

Recognition and service:

In 2026 I received an Anthropic Claude for Science award as sole Principal Investigator to support reliable and responsible speech AI for health research. I was an invited session chair at Interspeech 2026 for the "Audio Segmentation" and "Multimodal and Non-Speech Healthcare Applications" sessions. I am a member of IEEE (since 2017), the Australasian Speech Science and Technology Association (ASSTA) and the Speech and Language Biomarker Network, which began as the Harvard-MIT Speech and Language Biomarker Interest Group.

Research leadership and supervision:

Since January 2026 I have been Graduate Research Coordinator for the School of Engineering. I supervise higher degree research (HDR) candidates in speech AI, federated learning and biomedical signal processing.

Publications and impact:

My publications form a coherent research programme in AI-enabled speech assessment, biomedical signal processing, privacy-preserving health AI and deployable clinical machine-learning systems, with outputs in IEEE Transactions on Neural Systems and Rehabilitation Engineering, IEEE Internet of Things Journal, Nature Scientific Reports, Annals of Biomedical Engineering, Biomedical Signal Processing and Control, and International Journal of Machine Learning and Cybernetics. 
Across my five SciVal research areas, my FWCI exceeds the world average (FWCI >1), spanning Federated Learning for Data Privacy, Deep Learning for Anomaly Detection, ECG/Arrhythmia Detection, Speech Dynamics in Parkinson’s and Dysarthria, and Cerebellar Ataxia and Rehabilitation.

Research Interests:

• Speech signal processing and representation learning
• Advanced signal processing
• Human motion analytics
• Scalable AI in healthcare
• Quantum-inspired machine learning