Invited Speakers  特邀嘉宾

Invited Speakers


       Prof. Kunbao Cai
       Chongqing University, China


Prof. Kunbao Cai was born in Shanghai City of China, in 1950. He graduated from the Department of Electrical Engineering, Shanghai Jiao Tong University, China, in 1978, and received the M.Eng. degree in 1983 and the Ph.D. degree in 1998, all in electrical engineering, from Chongqing University, China.
From 1978 to 1980, he assisted teaching in the Electronic Department, Shanghai Jiao Tong University. From 1991 to 1994, he assisted researching on signal processing and system identification, and then studied, in the Department of Biomedical Engineering, McGill University, Canada. He held the following teaching and research positions in Chongqing University: Teaching Assistant, 1983-1986; Lecturer of Electrical Engineering, 1986-1994; Associate Professor of Electrical Engineering, 1994-2000; Full Professor of Electrical Engineering, from 2000. He retired from Chongqing University in 2015.
His major research interests include digital signal processing, biomedical signal processing, and artificial neural network with application to biomedical engineering. He is the author of textbooks: Digital Signal Processing (English Editions I and II), Publishing House of Electronics Industry, Beijing, China, (2007 and 2011). He completed a number of teaching research projects: Chongqing municipal excellent course of Signals and Linear Systems (from 2006), appointed as a Chief Prof.; Chongqing municipal excellent course of Digital Signal Processing (from 2008), appointed as a main Prof.; national bilingual teaching demonstration course of Signals & Systems (from 2010), appointed as a Chief Prof. Also he received a number of awards at Chongqing University for excellent teaching and textbook publications.

Speech Title: Transforms of multiwavelet and multiwavelet packet with application in identifying heroin addict pulse signals

Abstract: It is well known that the multiwavelets are a natural generalization of the scalar wavelets and can be viewed as vector-valued wavelets which have several advantages in comparison to scalar wavelets. Correspondingly, the multiwavelet transforms can provide more adaptive to satisfy the requirements of a variety of signal analysis. On the other hand, the multiwavelet transforms have also a fast computational structure for multiresolution analysis, which can be viewed as a generalization of the fast algorithm of Mallat’s multiresolution analysis for the case of scalar discrete wavelet transforms. However, while using the Mallat’s multiresolution analysis method to realize a multiwavelet transform, only the decomposed lowpass component at a decomposition stage is further decomposed, and the highpass component is left unchanging further. Therefore, the multiwavelet transforms can not supply finer time-frequency localized information for high frequency component obtained at every decomposition stage. Naturally, the concept of wavelet packet transforms was introduced into the multiwavelet transforms, which led to the so-called multiwavelet packet transforms. Thus, both the lowpass and highpass components at any decomposition stage can be further decomposed, with the result that any finer degree of the time-frequency localization can be obtained. It is really a great interesting to explore the effectiveness of these two modern signal processing techniques in identifying heroin addict pulse signals. In the research, the transforms of multiwavelet and multiwavelet packet are, respectively, used to decompose pulse signals collected from 15 heroin addicts and 15 healthy normal subjects. Combining entropy techniques in the feature extraction, the extracted feature vectors have good distributive properties in feature plane. To obtain a good generalization for classification of two classes of pulse signals, the support vector machine is introduced. It is expected to receive good research results.