Breaking Down Barriers: Advancing Interdisciplinary Speech Applications in Children’s Development

Jialu Li, February 28, 2024

Children’s development can evolve rapidly within the first few years of life. Identifying children at risk of developmental disorders or autism, as well as those with speech and language delays, and providing early intervention, is essential for improving their communication abilities, learning potential, and social interactions. My vision aims to harness cutting-edge AI speech technology to develop innovative screening solutions to early identify those with special needs at each developmental milestone.

In this talk, I will first present my work in advancing interdisciplinary speech applications to capture critical adult-child vocal interactions under different social contexts, including monitoring infant psychological development at home and detecting children who are at risk of autism at clinics. By leveraging the self-supervised learning scheme, the data sparsity issues can be significantly alleviated. Additionally, I will delve into a project that studies multilingual technology, specifically focusing on enhancing phoneme recognition in under-resourced Asian tonal languages through transfer learning from rich-resourced languages. This project examines the linguistic attributes of phones and tones, with the potential benefits of better understanding of language development or detecting pronunciation errors in multilingual children. Finally, I will conclude the talk by outlining my short-term and long-term research plans for integrating advanced children’s speech technology with conversational AI technology to develop innovative and interactive AI solutions for speech and language therapy.