Any speech recognition system is the building of a speaker independent system that recognizes in real-time, natural, fluent, spontaneous, continuous speech and which is capable of telling speech and background noise apart.In this regard, this book attempts to developed prototype speaker independent and continuous speech recognizer for Tigrigna language and investigate the basic units of recognition for Tigrigna and has identified the possible challengers as well.Besides, it has tried to model using Hidden Markov Model base speaker independent continuous phonemes recognition and it is based on pure acoustic phonetic information. Finally, Tigrigna phonemes recognizers¿ prototype is built for both speaker-dependent and speaker-independent models using these contending sub-word units. Last but not least, comparative analysis of these contending sub-word units has been made and also tested the performance of the system and compare its accuracy and recommended measures for similar research and projects.This work there for, will be useful to researchers, speech application developers and other educators or institutions working on similar projects and researches.