Computer control with Sphinx-4 speech recognition software

Denis Bučković (2013) Computer control with Sphinx-4 speech recognition software. EngD thesis.

Preview

PDF Download (813Kb)

Abstract

With the development of systems for recognizing human speech has been engaged people from academic and professional field for many years. Two of the systems for speech recognition as also Sphinx-4 and Julius, that in diploma work we compare each other. With the help of Sphinx-4, we created an application for the management of the computer in which the speaker sends sound signals to the microphone and with test tool we measured its efficiency in terms of the word error rate. At the beginning of the thesis is acquainted with human speech, which is the process of communication. Followed by extensive presentation of the general speech recognition system, where we introduced with the concepts, that in continue enable us to understand the Sphinx-4 and Julius system. The purpose of this chapter is presentation operation of the speech recognition system, show it’s strengths and weaknesses and what it restricts it or which problems it facing with recognizable natural human speech. Then acquainted with Sphinx-4 and Julius systems. At Sphinx-4 system we look the architecture, of which in the practical work serves for easier understanding functioning of the system, a little rougher, this we presented in the Julius system. Followed by a section, which compare each other the aforementioned system. Here we shows some strengths, weaknesses, differences and similarities between them and with testing we present the results of performance of the speech recognition Sphinx-4 system. The last part of the thesis we present the operation of applications and we show the user interface looks.

Actions (login required)