Využití Soft Computingových Metod pro zpracování řečového signálu v reálném prostředí Smart Home

Abstract

The aim of this diploma thesis is investigate the field of Soft Computing methods in Speech Signal Processing and find an appropriate method for noise suppression and use it in simulation and practical implementation. The chosen method will be theoretically and mathematically described and used in implementation to suppress the ambient noise from speech signal for controlling a Smart Home based on the KNX bus system. Voice control of the operational technical features of this communication bus system is a prerequisite for a simpler household management, eliminating otherwise necessary manual handling of the control device, especially for seniors or disabled persons. However, in households there are a number of disturbing elements, including for example the noise of household appliances or weather conditions that may cause this control to malfunction and noise need to be removed. Nowadays, a number of filters are already available to satisfactorily remove pre-specified noise, but the use of non-linear adaptive methods could shift this results to a completely different level. After successful implementation, it is necessary to evaluate the signal if the used noise suppression method has been successful.