演講題目(Topic)：From Digital Signal Processing to Machine Learning

演  講  者(Speaker)：Prof. Hen-Geul Yeh (California State University, Long Beach, CA, USA)

時　　間(Date & Time)：2019/7/16, Tuesday, 14:30~16:00

地　　點(Location)：清華大學台達館 304室 (R304, Delta Building, NTHU)

ABSTRACT

Digital Signal Processing (DSP) is a general technical term or function which means the processing of discrete-time signals or data sequence either in real-time or off line to get the desirable result. In this talk, we review the history of the development of DSP and the connection to the applications in Machine Learning (ML), including both hardware and software algorithms. Since World War II, if not earlier, electronics engineers have speculated on the applicability of digital hardware techniques to the many problem areas in which signal processing plays a role. Needless to say, the conclusion was not favorable. Cost, size, power consumption, and reliability strongly preferred analog filtering and analog spectrum analysis techniques. It was not until the mid 1960’s that a more formal theory of DSP began to emerge. By then the potential of integrated circuit technology was appreciated and was reasonable to complete signal processing systems that could best be synthesized with digital components. As of today, there are many embedded hardware platforms build on very large scale integrate circuits (VLSI), such as floating-point digital signal processors as well as field programmable gate array (FPGA). At the same time, DSP techniques have been advanced rapidly in recent years and have found many applications in almost every field of technology. In software algorithm development, it starts from adaptive signal processing (~1970), then moves to machine learning, neural network, and finally named as artificial intelligence, which becomes a rapid growing field today. In this talk, we apply ML techniques to classify shark behavior via a k-th nearest neighbors (kNN) method, which is one of the most used learning algorithms in industry. kNN is one of the simplest classification algorithm, no assumptions about data, relatively high accuracy, with the cost of computational load.

BIOGRAPHY

Hen-Geul Yeh received the B.S. degree in engineering science from National Chen Kung University, Taiwan, ROC, in 1978, and the M.S. degree in mechanical engineering and the Ph.D. degree in electrical engineering from the University of California, Irvine, in 1979 and 1982, respectively.

Since 1983, he has been with the Electrical Engineering department at California State University, Long Beach (CSULB), USA, and served as the department Chair since 2016. In addition to his technical and engineering excellence, he was selected as a NASA JPL Summer Faculty Fellow twice, in 1992 and 2003, respectively, and the Boeing Welliver Faculty Fellow in 2006. His research interests include DSP/Communication/Control algorithms development, and implementation using FPGA and digital signal processors with applications to communication systems, smart grids, controls, and smart systems, with focus on Wi-Fi, adaptive systems, and mobile communication in multipath fading channels. He has published more than 100 research papers on Signal Processing, Communications, Controls, and Smart Grids.

Dr. Yeh is a professional engineer in Electrical and is the recipient of five NASA Tech. Brief and New Technology awards from the National Aeronautics and Space Administration (NASA), the inventor’s award and other awards at the Aerospace Corporation, the Northrop Grumman Excellence in Teaching award, College of Engineering, CSULB, 2007, the Distinguished Faculty Scholarly and Creative Achievement Award, CSULB, 2009 and Outstanding Professor Award, CSULB, 2015. He has received four US patents in the area of Signal Processing, Communication and Controls. Since 2010, he has served as the organizer and Conference Chair of IEEE Green Energy and Smart Systems Conference (IGESSC).