Public examination of a doctoral dissertation in the field of computer science
Doctoral candidate: M.Sc. Rahim Saeidi
Date and venue: 31.5.2011, at 12 noon, M100 Metria Joensuu campus
In this dissertation Rahim Saeidi has studied speaker recognition in which the identity of a person inferred from her voice. Speaker recognition is getting widely used for remote access to computers and e-banking as well as forensic applications. Several aspects of this challenging task are addressed and new solutions are proposed. His thesis mainly investigates and proposes novel ideas for the questions below:
1- How to make the system to be robust against environmental noise?
2- How to recognize the identity of two speakers talking at the same time?
3- Is it possible to make the recognition system to respond faster in price of reduced recognition accuracy?
Conventional and novel parametric and non-parametric spectrum estimation techniques are considered in system front-end. Statistical modeling approach (Gaussian mixture model) is employed for speaker modeling and significance test is utilized to elaborate the results. The results indicate that more accurate spectrum estimation leads to better recognition accuracy. The proposed system for recognizing two speakers in a single-channel mixed-speech achieves an accuracy of 92% for recognizing both of the speakers. Faster recognition in the order of 5 achieved with only 1% loss in recognition accuracy.
The doctoral dissertation of M.Sc. Rahim Saeidi entitled “Advances in Front-end and Back-end for Speaker Recognition” will be examined at the Faculty of Science and Forestry. The opponent in the public examination is Professor Haizhou Li, Institute for Infocomm Research Singapore and the custos is Professor Pasi Fränti of the University of Eastern Finland.
Photo available for download at http://www.uef.fi/vaitoskuvat
Contact: Rahim Saeidi, firstname.lastname@example.org, tel. +358 44 591 2350
Publishing year: 2011Back to this years article listing