题目：A Discussion on Theoretical Foundation of Modern Speech Recognition Technology and Beyond
时间：4月11日下午14：00-16：00
地点：曹光彪大楼西楼205
主讲人：霍强博士 微软亚洲研究院
Abstract:
Modern automatic speech recognition (ASR) technology is based on a communication theoretic view of the generation, acquisition and transmission, and perception of speech. The goal of speech recognition is then defined as recovering the intended sequence of linguistic units from the observed acoustic signal. Under a statistical decision theoretic formulation, ASR is formulated as a statistical decision problem. In this talk, I will discuss the theoretical aspects of this formulation; explain why the current ASR approach works so well in certain conditions, and more importantly why it does not work in many other situations. During the process, I will clarify some important concepts about normalization, adaptation, and robust decision. I will also suggest how we can go beyond the current ASR paradigm based on the lessons I learned from doing OCR research in the past decade.
About the Speaker:
Dr. Qiang Huo joined MSRA on August 1, 2007 as a Lead Researcher in Speech Group. Prior to joining MSRA, Qiang had been a faculty member at the Department of Computer Science, The University of Hong Kong since 1998. From 1995 to 1997, he was with the ATR Interpreting Telecommunications Research Laboratories, Kyoto, Japan. From 1991 to 1994, he did his Ph.D. research at the Department of Computer Science, The University of Hong Kong. In the past 20 years, he has been doing research and making fundamental contributions in the areas of speech recognition, handwriting recognition, OCR, biometric-based user authentication, hardware design for digital signal processing, image processing and computer vision.
Qiang received the B.Eng. degree from University of Science and Technology of China (USTC), Hefei, China, in 1987, the M.Eng. degree from Zhejiang University, Hangzhou, China, in 1989, and the Ph.D. degree from the USTC, in 1994, all in electrical engineering.
本次报告计读书报告一次，请各位同学务必要签名，报告会开始后不再接受签名，在报告结束后2个工作日内（截止到4月15日）上班时间到曹光彪主楼217室贺老师盖读书报告章，过期作废。
欢迎广大师生踊跃参加。
计算机学院
2008-4-7
时间：4月11日下午14：00-16：00
地点：曹光彪大楼西楼205
主讲人：霍强博士 微软亚洲研究院
Abstract:
Modern automatic speech recognition (ASR) technology is based on a communication theoretic view of the generation, acquisition and transmission, and perception of speech. The goal of speech recognition is then defined as recovering the intended sequence of linguistic units from the observed acoustic signal. Under a statistical decision theoretic formulation, ASR is formulated as a statistical decision problem. In this talk, I will discuss the theoretical aspects of this formulation; explain why the current ASR approach works so well in certain conditions, and more importantly why it does not work in many other situations. During the process, I will clarify some important concepts about normalization, adaptation, and robust decision. I will also suggest how we can go beyond the current ASR paradigm based on the lessons I learned from doing OCR research in the past decade.
About the Speaker:
Dr. Qiang Huo joined MSRA on August 1, 2007 as a Lead Researcher in Speech Group. Prior to joining MSRA, Qiang had been a faculty member at the Department of Computer Science, The University of Hong Kong since 1998. From 1995 to 1997, he was with the ATR Interpreting Telecommunications Research Laboratories, Kyoto, Japan. From 1991 to 1994, he did his Ph.D. research at the Department of Computer Science, The University of Hong Kong. In the past 20 years, he has been doing research and making fundamental contributions in the areas of speech recognition, handwriting recognition, OCR, biometric-based user authentication, hardware design for digital signal processing, image processing and computer vision.
Qiang received the B.Eng. degree from University of Science and Technology of China (USTC), Hefei, China, in 1987, the M.Eng. degree from Zhejiang University, Hangzhou, China, in 1989, and the Ph.D. degree from the USTC, in 1994, all in electrical engineering.
本次报告计读书报告一次，请各位同学务必要签名，报告会开始后不再接受签名，在报告结束后2个工作日内（截止到4月15日）上班时间到曹光彪主楼217室贺老师盖读书报告章，过期作废。
欢迎广大师生踊跃参加。
计算机学院
2008-4-7
