Pamuditha Somarathne

I am a PhD student working at the intersection of human-centred AI and machine learning, with a focus on modelling human motion for accessibility. My research explores pose estimation and representation learning to predict, synthesise, and understand human motion in realistic settings. I am particularly fascinated by the complex kinematics of sign language, and I aim to build interactive technologies that accelerate sign language learning and support the Deaf community. I also maintain active research collaborations in biosignal processing (PPG/PCG) for health monitoring.

Now: PhD Candidate, The University of Sydney
Background: BSc in Electronic and Telecommunication Engineering, University of Moratuwa

Featured Projects

all projects →

SignPlay: Learning Australian Sign Language Through Interactive Mobile Games

An 'AI for Good' initiative empowering hearing individuals to learn Australian Sign Language (Auslan) using AI driven Apps.

A Dual Classifier-Regressor Architecture for Heart Sound Onset/Offset Detection

This project explored robust detection of heart sound onset and offset events by combining phonocardiogram (PCG) and electrocardiogram (ECG) signals using a joint classifier-regressor architecture.

Generative Motion Modelling

PhD research developing generative architectures for high-fidelity 3D human motion reconstruction and optimising latent space representations for sparse sensor data inputs.

Publications

all pubs →

A Dual Classifier-Regressor Architecture for Heart Sound Onset/Offset Detection

IEEE Transactions on Biomedical Engineering (TBME) 2026 (2026)

VideoPulse: Neonatal Heart Rate and Peripheral Capillary Oxygen Saturation (SpO2) Estimation from Contact Free Video

arXiv preprint arXiv:2602.23771 (2026)

Latest News

all news →

VideoPulse preprint released

Feb 01, 2026 - Preprint on neonatal heart rate and SpO2 estimation from contact-free video is now on arXiv.

Heart sound segmentation paper in IEEE TBME

Jan 15, 2026 - Our dual classifier-regressor architecture for heart sound onset/offset detection is out in IEEE Transactions on Biomedical Engineering (early access).