Return Home

Speech Enhancement Using Microphone Arrays

Speech enhancement in adverse environments remains a very challenging research problem. This project aims to use microphone arrays to bring speech enhancement to a greater level of performance. Here, two approaches will be investigated, namely beamforming and blind signal separation (BSS). Specifically, the objective is to improve the performance of both approaches in high noise and high reverberant environments. Moreover, this project will draw the expertise of the group on beamforming to integrate with BSS, i.e., synergetically combine beamforming and BSS.

Some specific objectives are to establish and understand the performance bounds on beamforming and BSS, ways to combine them, address convergence issues and to incorporate real-time constraints on development of the algorithms. Additionally, post-filtering or single channel techniques will be investigated. Previous studies show that post-filtering techniques play a vital role for both BSS and beamforming as they can be designed to increase further the level of noise suppression. Also it is important to study the behaviour of the above stated algorithms in a real time environment.

Research topics

• Adaptive Beamforming, ABF
• Blind Signal Separation, BSS
• Single-channel/Post-Filtering, PF
• Realtime implementation of acoustic signal processing, RTI