International Conference on Knowledge Management

This paper will report on a research to develop Speaker Independent, Continuous Speech Recognizer for Kafi Noonoo (Afro-Asiatic language that belongs to North Omotic sub family in Ethiopia) using Hidden Markov Modeling technique. The portable and open source toolkit called Hidden Markov Model (HMM) Toolkit is used to perform the experiment. The development of HMM based Automatic Speech Recognition (ASR) requires both text and speech corpus for training and testing the HMM. In order to have a model that incorporates different features of the language, we included the different dialects of Kafi Noonoo in the corpus and then prepared the training and test corpus from the scratch, and after preprocessing we have sampled and performed feature extraction using Mel Frequency Cepstral Coefficients (MFCC) feature extraction technique