A Continuous, Speaker Independent Speech Recognizer for Afaan Oroomoo

This thesis work is performed through integration and implementation of appropriate tools, techniques and methodologies. Furthermore, the special features of the language are also entertained due to the speech characteristics taken in to considerations, so as to arrive in to the appropriate results for the language under investigation. Mainly the purpose of this research was towards exploring the possibility and applicability of the selected tools and techniques to continuous Afaan Oromoo speech utterances and the possible outcomes that recognizer brought up. To arrive to this, the researcher performed different activities towards achieving the objective. Accordingly the open source tool from Sphinx System, Sphinx Train and Sphinx4, was selected which work with the statistical approach under HMM for continuous Afaan Oromoo speech recognizer as the tool have a variety of modules that implement state-of-the-art speech recognition techniques beyond its modularity and flexibility. As to the knowledge of the researcher, this research is the first of its kind for continuous Afaan Oromoo speech recognition where as one attempt for isolated word recognition for Afaan Oromoo made earlier.

Autorentext

Kassahun Gelana received his Bachelor of Science degree (BSc) in Information Technology from Jimma University, Ethiopia (2006) MSc degree in Information Science from Addis Ababa University, Ethiopia (2010). Currently he is Lecturer at Jigjiga University, Under Faculty of Engineering and Technology, Department of Computer Science, Ethiopia.