Thank you to the following supporters of ICASSP 2007
Dr. Amitav Das, Microsoft Research, India
Biometric recognition, or, the automatic recognition of individuals based on their biometric features, is gaining increasingly more and more attention and enjoying increasing demands in various access control and crime/terrorism prevention usages. Compared to the use of a single biometric, the use multiples biometrics offers higher performance as well as greater robustness to sensor failures and adverse background conditions. This tutorial will focus on audio-visual biometric recognition, which uses both speech and face images as biometrics. This is at present one of the most promising bimodal biometric recognition method, which is setting trends in delivering reliable high performance in adverse background conditions. Audio-visual biometric recognition is also highly robust in imposter attack and “aliveness” tests as combination of both, especially in live video format, prevents attacks using recorded speech and/or images.
The sensors for audio-visual biometric recognition also are everywhere on the globe and becoming cheaper and “mobile” as most of the mobile phones are being equipped with a camera. Laptops are coming with integrated web-cam, thereby heralding the advent of “bio-log-in” and on-line any-time/any-where user authentication using audio-visual recognition methods. In emerging economies such as India, such Audio-visual biometric authentication is expected to be a key enabler to allow a vast number of illiterate/semi-literate customers to gain access to various services. In addition to facilitate commercial transactions, AV authentication also allows next generation phone and PCs to offer a more customized features and operation modality tailored to the user’s needs and usage patterns. We envisage audio visual biometric recognition to become an integral component of the next-generation human machine interface (HMI).
This tutorial will focus on the various core technologies needed to build a good audio-visual biometric system and also review various applications and related issues and requirements. After some introductory definitions and basics of biometric recognition, the main body of the tutorial will cover speaker recognition, face recognition and various fusion methods in details, followed by a detailed review of recent research results in audio-visual biometric authentication. The review in each case will contain coverage of the state of the art methods as well as some of our own recent research results at Microsoft Research. Finally some emerging trends and applications will be presented. The outline of the tutorial is given below:
The material presented will build up from basics to complex theories with lucid examples and should be of interest to students, researchers, and engineers who are interested or working in biometrics and above mentioned applications. Basic knowledge of probabilities and signal and image processing will be helpful.
A complete list of over 200+ relevant papers will be presented along with the tutorial handout
Dr. Amitav Das [Ph.D.(1996) Computer & Systems Engineering - University of California – Santa Barbara; M.S. (1986) Computer Engineering, Rensselaer Polytechnic, Troy, New York; B.Tech. (1984) Electronics & Telecom. Engineering, Jadavpur University, Calcutta, India] had rich experience with hands-on-research as well as managing core research and product development teams in various leading companies around the globe such as Philips Research Panasonic Research, Qualcomm, Motorola, Siemens Research and currently Microsoft Research, India. During his 20 years plus career span in the area of signal processing, he had contributed towards many products and solutions as well as various standards in international standard bodies such as ITU-T study group 16, IETF, 3GPP/3GPP2, US TIA/EIA, etc. Dr. Das has 12 US and world-wide patents granted and few recent invention disclosures and Indian patents filed and he had several (24+) publications and tutorials.
At present, Biometric Recognition is one of his key research areas. Several novel methods are being investigated by him under various new applications domains. In 2006, he co-presented a Tutorial in Speaker Recognition in ICASSP, which was well received. He published several papers recently in ICASSP-06 and Odyssey-06. Several other papers exploring new methods and applications of Biometric Recognition, namely speaker recognition, face recognition and audio-visual biometric recognition, are underway. He has given expert talks and seminars in various national meets of experts in this area. His other research interests include human-machine interface, signal compression & coding, pattern recognition and machine learning techniques, embedded signal processing. Dr. Das is a member of IEEE, ACM, and other professional organizations and he is an adjunct faculty of the Indian Institute of Information Technology (IIIT-Bangalore), an emerging premier institution in India.