2DN Undergraduate Students Won 2026 Image and Voice Processing Award

Jayde Lal and Jack Matthews, two 2DN undergraduate studetns, won the 2026 best undergraduate research project award in the Image and Voice Processing category. In this project, Jayde investigated how reinforcement learning algorithms improve the performance of Automatic Speech Recognition for dysarthric speakers. She conducted a comprehensive investigation and benchmarked the PPO, GRPO, DPO, and the recently proposed GRDPO algorithms across three tasks: error detection and localisation in dysarthric transcripts, spoken keyword classification, and sequential decision-making in Blackjack. Moreover, Jack studied post-processing techniques for error correction and n-best reranking, and implemented a novel post-processing method that incorporates user input to improve ASR outcomes further when presented with dysarthric speech. 

This award was presented to Jayde and Jack by independent judges from Garmin New Zealand. The DeepNet Discovery Network congratulates them on this accomplishment.