During the Spring 2008 semester I’ll be teaching a graduate course (900 level) on discrete time speech signal processing. This course will cover topics on the theory and application of discrete time speech signal processing. First we’ll discuss the speech production mechanism and anatomy, as well as linear models of speech production. Next, we’ll delve into speech signal analysis and synthesis methods: pole-zero speech models, homomorphic processing, short-time Fourier transform processing, filter banks and sinusoidal synthesis and analysis. We’ll continue with a look at frequency domain pitch and voicing estimation. These topics are of particular interest to me these days, because one of my PhD students, Alex Shyrokov, is using prosody in modelling human-human spoken dialogues, and another PhD student, Zeljko Medenica, may use prosody in detecting frustration in drivers.
The course will conclude with a look at applications of speech processing, with topics selected from the following areas: speech modification, speech coding, speech enhancement and speaker recognition. Note that some very prominent applications of speech processing, such as speech recognition and text-to-speech converstion will not be covered in this course. These topics would require deeper coverage of statistical discrete-time signal processing.
I will use a book by Thomas Quatieri, called Discrete-Time Speech Signal Processing. The book has many Matlab problems, which will be very helpful. It also has a similar feel to my favorite DSP book, Oppenheim and Schafer’s Discrete Time Signal Processing. I’ve used the Oppenheim and Schafer book as the source of many notes for the Introduction to DSP course (ECE 714/814) at UHN that I’ve taught several times.
I’m looking forward to teaching this course. Feel free to email me if you have questions or comments (andrew dot kun at unh dot edu).
Andrew Kun