Developing a speech recognition system for the Sinhala language | |
| Author | Pragnaratna, Mudunkotuwage Sidath |
| Call Number | AIT Thesis no.ISE-07-26 |
| Subject(s) | Speech perception Sinhalese language |
| Note | A thesis submitted in partial fulfilment of the requirements for the degree of Master of Engineering in Mechatronics, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. ISE-07-26 |
| Abstract | There are fully-developed speech recognition systems available for the English and other languages today. These systems help to solve great number of human problems in our day-to-day life. However, there is not any speech recognition system available for the Sinhala language so far. I developed a speaker-independent and single-word speech recog¬nition system for the Sinhala language, which mainly focuses people who have difficulties to use the computer keyboard for email writing. The developed system is a computer¬based dictation system for Sinhala email writing. I use a Hidden Markov Model (HMM) based approach to recognize unknown spoken utterances. To build the Speech Recogni¬tion System (SRS), I used HTK which is a free and open-source development toolkit for building experimental speech recognition systems. The size of the training vocabulary is 6688 words, most of which were taken from an editorial of a Sinhala newspaper for two months. The training corpus consisted of 1721 continuously uttered speech waveforms, on average 15 words per each sentence, recorded from 16 male and 13 female native Sinhala speakers at AlT. The SRS's performance was evaluated for the original training data. It shows more than 60% overall recognition accuracy for the recorded speech at word level for all 6688 words in the training data. However, the phoneme-level performance analysis shows promising results with almost 90% recognition accuracy. The computer based email dictation system was developed using ATK, which is an Application Program Interface (API) for HTK, provides the Graphical User Interface (GUI) to the underlying speech recognizer. The software application is for personal computers which uses Linux operating system, named as SinMail, will be free and open source for anybody to use it or to develop it further |
| Year | 2007 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. ISE-07-26 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Industrial Systems Engineering (DISE) |
| Academic Program/FoS | Industrial Systems Engineering (ISE) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Manukid Parnichkun;Rajatheva, R.M.A.P.; |
| Scholarship Donor(s) | Thailand (HM King); |
| Degree | Thesis (M.Eng.) - Asian Institute of Technology, 2007 |