Speaker diarization in meetings domain

Abstract

The purpose of this study is to develop robust techniques for speaker segmentation and clustering with focus on meetings domain. The techniques examined can however be applied to any other domains such as telephone and broadcast news. Traditional techniques for speaker diarization developed for telephone conversations or broadcast news are based on a single channel, which is notably different from meetings domain which can have multiple channels. These techniques when adapted to meetings domain however perform poorer than expected since they do not exploit direction of arrival information, which is available in many meeting rooms with the presence of multiple microphones. Moreover, many of these techniques are involved with tunable parameters, which are presumably derived using external data. These parameters need to be individually adjusted for each data set accordingly to obtain reasonable performance. In this thesis, the focus is on robust and accurate speaker diarization techniques in meetings. I also want to thank my colleagues Ma Bin and Kong Aik for their understanding and support at work to let me spend time pursuing my own research. Also to my friends Wang Lei and Xiao Xiong whose discussions though not many but are really enjoyable and helpful. To Hanwu and Tin Lay who contributed to the research platform which I am tremendously benefited from. Lastly, this work could not have been achieved without the love and support of my family, my dad who instills into me the persistence, my mum who cares for me anytime and my fiancee who loves me and brightens up my life.