Lecturers‎ > ‎

Michael Wagner

Michael Wagner

Faculty of Information Sciences and Engineering

University of Canberra


Professor Michael Wagner has a Master degree in physics from the University of Munich and a PhD in computer science from the Australian National University. His PhD thesis topic was a learning technique for speaker characteristics in continuous speech (1973). He has written over 160 papers in the fields of speech analysis, automatic speech recognition, and speaker verification. He is a Fellow of the Institution of Engineers Australia (EA) and a Senior Member of the Institute of Electrical and Electronic Engineers (IEEE). Dr Wagner was the Foundation President of the Australasian Speech Science and Technology Association (ASSTA) from 1988 to 1992, and he was a member of the Board of the International Speech Communication Association (ISCA) from 1999 to 2007. He is currently the chair of ASSTA’s Spoken Language Databases subcommittee and a member of its Forensic Speaker Identification subcommittee. He is also a member of the Editorial Board of the IET Journal on Biometrics. He is currently the Professor for Computing at the University of Canberra and Director of the Human-Centred Computing Laboratory at UC.

Talk : Speaker Recognition

Speaker recognition comes in two main varieties: firstly for the purpose of authenticating clients for telephone access to banking and similar services, and secondly for the purpose of providing evidence in forensic cases where voice recordings exist in relation to a crime. Speaker verification and forensic voice comparison largely share similar signal processing of speech signals, but differ somewhat in their feature extraction and statistical methodologies. This lecture will give an overview over the state of the art of speaker recognition, including the signal processing and feature extraction for speech signals, the statistical methods of scoring and decision making, and recent advances in coping with channel variations and environmental noise. Tutorial material will cover a typical speaker recognition task.