Digital Speech Processing – Content
In this two term lecture the fundamental issues, standards, and recent trends in speech communication technology are presented. The first part (DSV I) deals with the fundamentals of speech production, hearing and psycho-acoustic aspects. Further main topics are:
- Spectral transformations
- Filter banks for spectral analysis and synthesis
- Stochastic signals and estimation
- Linear prediction
- Quantization.
The second part (DSV II) covers the more application-oriented topics:
- Speech coding: model-based predictive coding, vocoder, adaptive quantization, adaptive prediction in the time domain, adaptive coding in the frequency domain, subjective speech quality, standards, realization aspects
- Error concealment and soft decision source decoding, reduction of annoying effects due to residual bit errors
- Bandwidth extension of speech signals
- Single- and multichannel noise reduction: reduction of acoustic noise signals by adaptive filtering, active compensation of acoustic noise, reduction of ambient reverberation
- Compensation of acoustic echo signals: digital loudspeaking telephone, acoustic human-machine interface
The aim of this lecture is to teach on one hand-side the fundamental techniques and algorithms of digital speech processing. On the other hand-side an overview of the state-of-the-art technology, actual applications and standards is presented.
Exercises are offered for both parts of the course.
Particular topics can be examined at the laboratories "Communications Engineering", "Digital Signal Processing", and "Matlab laboratory" in experiments which are coordinated with this lecture.
The lecure relies on the book: "Digital Speech Transmission", P. Vary, R. Martin, Wiley 2006.