Voice Conversion Using LSF Modification Algorithm in Harmonic and Noise Speech Model of Acoustical Processor in Embedded TTS

Abstract

This article presents new approach to embedded Text To Speech system. We present algorithms for TTS synthesis based on Harmonic and Noise speech model. Acoustical Processor, as a part of speech synthesizer, for voice conversion is presented. Speech parameters are modified in Acoustical Processor module , in order to voice conversion codebooks. Offline text dependent voice conversion algorithm creates target voice database codebook using analysis fragment of recorded speech. It allow to quickly adapt the TTS system speaker voice to the preferences of the user. Due to significant speech database minimalization system is adequate to implement in mobile devices, for sms message reading, eBooks reading, ect..