Subspace and Multitaper Methods for Speech Enhancement Approved by Supervisory Committee: Subspace and Multitaper Methods for Speech Enhancement

ACKNOWLEDGEMENTS I would like to thank Dr. Philipos C. Loizou, my advisor, for his many suggestions and constant support during this dissertation work. I am very grateful for those research opportunities he provided me in the past three years, and I believe this research experience will greatly benefit my career in the future. Among many things I learned from Dr. Loizou, persistent effort is undoubtedly the most important one, which enables me to sort out several important issues in the relatively old area of speech enhancement. I believe our work in the past several years advanced the state of art of the speech enhancement problem. My many many thanks to him! I am also very thankful to my committee members Dr. Nasser Kehtarnavaz, Dr. Aria Nosratinia and Dr. Murat Torlak, for their precious time and those valuable comments and suggestions in my thesis proposal defense, which dramatically improve my final dissertation. I took two courses taught by Dr. Kehtarnavaz who is a great teacher and very good at clarifying confusing concepts, and I learned a lot of things in his class. I should also mention that Dr. Nosratinia had several very helpful discussions about wavelet with us in our early days to explore the applications of wavelet theory to speech enhancement. It is well known in our department that Dr. Torlak selflessly serves as the current program chair of IEEE Signal Processing Society Colloquia and organized a lot of wonderful lectures, which widens a lot of people's knowledge and of course mine as well. Dr. Dian Zhou expressed his interests in my research work, and supplied me with his insightful perspectives about spline wavelet. Those ideas help me see the possible connections between the numerical methods for circuit simulation and the digital signal v processing theory. Dr. Saquib often talked with me about my research and gave me a lot of encouragement, his Adaptive Signal Processing course taught me a lot of matrix computation skills. I also want to thank Dr. Zhongnong Jiang at Texas Instruments for his friendship. I had the pleasure of working with my lab mates Arun, Bin, Felipe, Kalyan and Sundar. My friends Hua and Ruiming also gave me a lot of help in my PhD studies. They are wonderful people. I should also acknowledge that my PhD studies were supported in part by National Institutes of Health. Of course, I am so grateful to my …