Hidden Markov Model based Amharic Speech Synthesizer

Authors: Yitagessu Birhanu, Guntram Strecha, Rüdiger Hoffmann

Abstract:

This paper explains initial results of a research work on HMM based Amharic synthesis system using phonemes as the fundamental acoustic unit. A speech corpus uttered by a single speaker, and consisting of all diphones of the Amharic language is used for training. It is recorded at 16 kHz sampling rate and 16 bit sample depth, in a standard acoustic studio. Initial results show that the phoneme HMM based Amharic speech synthesizer has acceptable intelligibility and naturalness.