Syntéza zvukového signálu

Abstract

This bachelor's work deals with speech synthesis in frequency domain in Czech language. The resulting solution consists of speech analyzer and synthesizer of chosen elements of speech. Speech synthesis is based on LPC model. Speech analyzer analyses input speech element in WAV format, divides it into microsegments of length 20 ms and keeps LPC coefficients, voiced/unvoiced sign, and residual prediction error signal. Resulting data are saved in XML format into a file. Furthermore, methods of micro-segment analysis are mentioned, like: voice detection, base frequency detection. In domain of resulting speech synthesis are taken conversion of micro-segments back into time domain, phonetic transcription, dividing sentences into syllables and their concatenation.