Sphinx based speech recognition application for Sidama language

Abstract

Sidamic (sidaam-u afo in Sidama) is one of the languages from the East Highland Cushitic language family. It is a branch of Afro-Asiatic language phylum, spoken in the Sidama zone and some parts of the Southern Nations Nationalities and People's Region of Ethiopia. Although speech recognition systems can achieve accuracy of more than 95% for the read speech inputs, but this accuracy drastically decreases for the spontaneous speech inputs. On the other hand, speech recognition systems applied for real applications are required to be robust and expected to have high accuracy for spontaneous speech inputs. Human beings usually use multimodal communication when they speak with each other especially in noisy environments. In the starting of the last decade, the researchers started working on the spontaneous, robust, multimodal speech recognition systems. Nowadays speech recognition system has become a mandatory part of modern computing devices such as mobile phones, tablet devices etc. In this research, we have developed a spontaneous input speech recognition system for Sidama language using the Hidden Markov Model (HMM).