Course Contents
We begin the lecture series with a general introduction to the aspects of adaptive filters and possible generic applications. The goal is, among other things, to gain insight into when adaptive filters should be used advantageously.
Optimal filters, such as the Wiener filter, define the adaptation goal of adaptive filters. Therefore, we will first familiarize ourselves with them as a basis for the following derivations of adaptive filters. Another important concept, closely related to the Wiener filter, is the prediction filter concept, which will also be presented in the lecture. It is an important concept and, as such, forms the basis for various applications such as source coding or parametric filter design, to name just a few widespread applications.
In the following lecture unit, we will take a closer look at various adaptive filter concepts, including Least Mean Square (LMS), Recursive Least Squares (RLS), and Affine Projection (AP) algorithms. We will learn about their specific properties in terms of computational complexity, adaptation speed, and robustness. Good adaptation performance in noisy environments requires appropriate control of the adaptive filters. We will learn about optimal methods for this.
Another important aspect is optimal filtering in the state space domain. After an introduction to this concept, we will first learn about the Kalman filter concept, followed by the particle filter concept. Both will be related to their specific applications. We will also relate the state-space concepts to the LMS concept to understand their individual properties.
Deep Neural Networks (DNNs) are becoming increasingly important for AI applications. Their training also requires adaptive methods. This lecture will introduce the theoretical foundations and the concept of adaptive training of DNNs.
All theoretical concepts are explained using concrete examples – many, but not exclusively, based on audio signal processing. Four tutorials with programming applications are also part of the lecture concept. This allows for practical application of the course content and ensures a deeper understanding of the algorithms and their application to real-world problems.
Literature
Slides of the lecture.
Literature:
[list]
[*]E. Hänsler, G. Schmidt: Acoustic Echo and Noise Control, Wiley, 2004 (Textbook of this course);
[*]S. Haykin: Adaptive Filter Theory, Prentice Hall, 2002;
[*]A. Sayed: Fundamentals of Adaptive Filtering, Wiley, 2004;
[*]P. Vary, U. Heute, W. Hess: Digitale Sprachsignalverarbeitung, Teubner, 1998 (in German)
[/list]
Preconditions
Digital Signal Processing
- Lecturer: Henning Puder