Kinship Recognition Using Voice: Developing models for predicting blood relations from speech using phonetic, prosodic, and SSL-based embeddings.
Explainable Deepfake Detection and Source Tracing: Building interpretable anti-spoofing systems to help forensic experts understand "why" a speech utterance is classified as genuine or fake.
Spoken Language Diarization in Parliamentary / Courtroom Scenarios: Addressing intra-language speaker switching, overlapping speech, and unknown-language segments in real-world legislative and judicial recordings.
Dementia Detection and Assessment using Speech: Leveraging acoustic biomarkers, signal processing, and speech foundation models to characterize cognitive decline and support early screening.
Multilingual Spoof-Aware Speaker Verification: Creating a benchmarking framework for Indian languages using state-of-the-art TTS and VC systems to simulate diverse spoofing conditions.
ASR Fairness for Second-Language Speakers: Investigating bias in modern ASR systems for non-native speakers and applying fairness-driven fine-tuning strategies for equitable performance.