Adaptive filters and machine learning: applications on image, communications, and speech

Abstract

The equalization of communication channels is a subject very explored in the literature. However, when it is applied to chaos-based communication systems has many open questions. The aim of this project is to find ways to make the performance of these systems closer to that of conventional communication systems. A scheme that allows the switching between these two systems will be studied.We will address other applications that use adaptive filters such as blind image restoration by means of equalization algorithms, such as the convex combination between a blind algorithm and an algorithm in the decision directed mode.The combination will also be used to improve multikernel adaptive filtering. Finally, we intend to study machine learning techniques and apply them to nonlinear problems such as the voice activity detection. (AU)