Microsoft Word - wseas-paper.rtf

This paper presents a method for speech synthesis by using the generalized harmonic model(GHM) analysis. The aim of this technique to improve the accuracy of analysis parameters is derived from a looping frame harmonic technique. The estimates of harmonic amplitudes and phases are formulated as the peaks of the spectrum amplitude and phase witch located in the pitch frequency multiples exactly. In this idea, we use reliable point detection algorithm to select one period of signal. By repeating this frame, a periodic sequence is produced. Now, we are able to estimate the exact spectral parameters of the frame which results in the high quality parameter extraction. This method can work under the different frequency sampling rates (from 8 kHz to 44.1 kHz). Therefore this method has many applications in high quality tasks. Key-Words: Speech analysis synthesis – Harmonic model – looping frame