Large Vocabulary Spontaneous Speech Recognition For Tigrigna

Abstract

Speech is the most natural form of human communication and speech processing has been one of the most exciting areas of the signal processing. Speech recognition technology has made it possible for computer to follow human voice commands and understand human languages. Tigrigna is a widely spoken languages in Northern part of Ethiopia and an official language of Eritrea. In spite of its importance, research on Tigrigna Speech Recognition (SR) is unfortunately still very limited. This thesis proposes and describes a research attempt on designing and developing a speaker-independent Spontaneous automatic Speech Recognition System for Tigrigna. The acoustic model of the Speech Recognition System is developed using the Carnegie Mellon University’s ASR development tool (Sphinx) while SRIM tool is used for the development of the language model. The system uses a database containing 3,524 sentences and phonetic dictionary containing about 10731 words and their pronunciations among these 349 sentence were used for testing the performance of the system. For the Language Model 30937 sentences have been used to develop a trigram language that contains 57,847 unigram, 292554 bi-grams and 74,178 tri-grams. Performance tests were then conducted at various stages using test data and finally, 36.83% word level accuracy and 13.67% sentence level accuracy were obtained. The results are encouraging and with more optimization works better results can be achieved. Finally, conclusions were drawn and recommendations were made in line with the analysis and findings