Towards Next-generation Artificial Auditory System with Brain-inspired Technologies

Professor Xu, Dong   (Co-Principal Investigator (Co-PI) (for projects led by other university))

Tan Kay Chen   (Co-Investigator)

36

2024-06-30

700000

Towards Next-generation Artificial Auditory System with Brain-inspired Technologies

""1) Deep Learning"", ""2) Machine Learning"", ""3) Artificial Intelligence"", ""4) Target Speaker Extraction"", ""5) Artificial Auditory System""

Artificial Intelligence and Machine learning

Engineering (E)

C5052-23G

Collaborative Research Fund (CRF) - Group Research Project 2023/2024

2024

On-going

1. Develop a brain-like artificial auditory system that can emulate the biological counterpart across functional, structural, and mechanistic levels. The proposed system comprises a brain derived neural structure, along with faithful neuronal and network dynamics that can accurately reflect those found in the biological system. Furthermore, the system incorporates biologically plausible neural plasticity that can explain and analyze the development process of the auditory system.2. Develop a neural-steered target speaker extraction system that can replicate the selective auditory attention capability of humans. The system includes an EEG-based human-machine auditory interface that decodes auditory attention signals from real-time EEG recordings, along with a target speaker extraction model that utilizes the decoded attention signal to extract a sound source of interest from competing sounds. Additionally, active learning algorithms with user-friendly learning procedures will be developed to promote personalization and adaptivity to different listening environments.3. Develop a neural network-based model for binaural speech enhancement that can mitigate the distortion of binaural cues caused by traditional beamforming algorithms. Implement the proposed neural-steered target speaker extraction and binaural speech enhancement models on a binaural hearing aid research prototype. Conduct comprehensive experiments on individuals with normal hearing and hearing impairments to validate the effectiveness of the methods and models developed in this project.