Potential Future Research Directions

Abstract

The speech enhancement research presented here was motivated by several factors. Firstly, the development in recent years of audio-only hearing aids that utilise sophisticated decision rules to determine the appropriate level of speech processing served as an inspiration. A second motivational factor was the exploitation of the established cognitive relationship between audio and visual elements of speech to produce multimodal speech filtering systems. Another motivation was the desire to utilise audiovisual speech filtering to extend the concept of audio-only speech processing to become multimodal, from the perspective of potential application to hearing aids. Based on these motivations, the goal of the work presented in this book was primarily to develop a flexible two-stage multimodal speech enhancement system, working towards the development of a cognitively inspired fuzzy logic based speech enhancement framework that is autonomous, adaptive, and context aware. The novel proof of concept framework presented in this book makes use of audio-only beamforming, visually derived Wiener filtering, state-of-the-art lip tracking with Viola-Jones ROI detection, and a fuzzy logic controller, to present a novel speech enhancement framework. This chapter presents some conclusions and future work.