Course in Bachelor’s program (4th year, series E)

Teachers

Prof. Dragoș Burileanu
Lect. Șerban Mihalache

Course Description

This subject is studied within the field of Electronic Engineering, Telecommunications and Information Technologies / specialization Microelectronics, optoelectronics and nanotechnologies, and aims to familiarize students with the several advanced topics in the field of digital signal processing (statistical processing of random signals, spectral analysis, adaptive filtering, multirate signal processing, neural networks, and machine learning methods), with applications in communication, speech technology, and audio processing. The objective is to understand the phenomena underlying the studied techniques and their implementation in real systems, as well as to introduce modern signal processor architectures and their use in real-time processing systems. The numerous examples and detailed explanations given in the lecture notes and chapters help both to clarify more difficult theoretical aspects and to solve practical applications and problems, relevant for engaging the students in the learning process. Additionally, the laboratory applications have as objective acquiring practical skills related to the key theoretical concepts taught in class. The applications include various software simulations using a high-level programming environment (MATLAB).

The subject addresses the following basic ideas and specific concepts: discrete random signals and the response of digital filters to random signals, non-parametric and parametric methods of spectral analysis, random signal modeling, linear estimation and the Wiener filter, adaptive filtering, multirate signal processing, speech signal analysis and processing, digital processing techniques for audio applications, the use of artificial neural networks in signal processing, digital signal processors for real DSP applications. All these contribute to providing students with an overview of the methodological and procedural benchmarks related to the DSP field.

Contents

Course

• “Statistical signal processing” – Introduction. Continuous-time random signals; basic statistical parameters. Discrete-time random signals; the Wiener–Khintchine theorem. Digital filter response to random signals
• “Spectral analysis and parametric estimation for random signals” – Spectral analysis: nonparametric power spectrum estimation; signal modeling and parametric spectral estimation. Optimization algorithms. Linear estimation; Wiener filters
• “Adaptive filters” – Basic concepts. Adaptive algorithms: LMS, NLMS. Adaptive filter configurations: system identification, inverse modeling, linear prediction, interference cancelation; applications. Acoustic echo cancelation in distance talking communication systems
• “Multirate signal processing” – Generalities. Decimation by an integer factor. Interpolation by an integer factor. Sampling rate conversion by a rational factor
• “Speech analysis and processing” – Speech production and perception. Acoustic and phonetic level descriptions; representations in the time and frequency domains. Speech signal variability. The principle of linear prediction in speech technology; the LPC vocoder. Other applications of speech technology: intelligent dialogue systems; speaker recognition; forensic expertise of speech in audio recordings
• “Digital processing techniques and signal processors for audio applications” – Digital recording / playback systems on compact disk. Digital processing techniques used in professional audio studios: volume control and mixing, dynamic range modification, filtering and equalization, special effects. Audio digital techniques implementation using signal processors
• “Artificial neural networks and machine learning. Applications in signal processing” – Introduction; using neural networks in signal processing. Main features and patterns; learning principles. Adaptive processing using neural networks; nonlinear adaptive systems. Artificial intelligence; machine learning techniques. Neural networks – present and future trends

Laboratory

• Discrete-time deterministic signals: FFT, digital filters (MATLAB review). Discrete-time random signals: representation, statistical parameters
• Spectral analysis for random signals. Linear estimation; the Wiener filter
• Adaptive filters. The LMS and NLMS algorithms; applications
• Multirate signal processing: decimation, interpolation, resampling by rational factors; applications
• DSP techniques for audio and speech processing applications
• Use of adaptive filtering for speech enhancement
• Laboratory assessment

Grading

Laboratory evaluation (colloquium): 25%
Course midterm exam (written): 25%
Course final exam (written): 50%