In this talk, Prof. Ming will discuss the fundamentals of automatic speech recognition (ASR), including the ‘classical’ approaches, and the most recent advances including the deep-learning based approaches. He will focus on the search of the solutions to a major and unsolved challenge – noise. Noise is unavoidable in many real-world applications, and the current ASR methods lack noise robustness. The talk will introduce the state-of-the-art approaches for improving the noise robustness, their shortcomings, and some of our recent advances towards an improved solution.
Ji Ming is a Professor in EEECS at QUB. His research interests are mainly in audio, speech and language processing. He was a Visiting Scientist at the MIT between 2005-2006. Prof. Ming and his team have produced the widely-cited approaches for speaker recognition in noise, and the award-winning systems for speech separation and speech enhancement.