Pokročilé metody kompenzace šumu pro rozpoznávače řeči na bázi virtuální instrumentace

Abstract

This thesis deals with advanced noise compensation methods (environment noise) for speech recognition based on virtual instrumentation. The primary goal of this thesis is to design and implement a virtual instrument for testing selected methods of noise suppression designed to improve the resulting speech recognition success. The theoretical part of thesis is devoted to the literature review of the current state of the issue of noise reduction for speech recognizers. The experimental part describes the design and implementation of a virtual instrument for automated testing of selected noise suppression methods. In the experimental part was created a test database of real recordings to perform a series of experiments. Evaluation of quality of filtration is performed on speech recognition success rate and signal to noise ratio.