원문정보
초록
영어
This paper describes a small vocabulary Bengali database development to evaluate the performance of speech recognition algorithms in clean conditions. The database is constructed by Bangla digit sequences (/ak/, /dui/, /tin/, /chaar/, /panch/, /chhoy/, /shaat/, /aat/, /noy/, /zero/, /shunno/) are used. The developed database is consisted of two sets of data which are training and testing datasets. The training dataset contains 3824 utterances of 50 speakers; on the other hand, the testing dataset is subdivided into four groups (clean1, clean2, clean3 and clean4) and contains 1985 utterances of 52 speakers. In both sets of data the speaker’s age ranges from 19 to 25 years. All the recordings have been done in a quiet room but not soundproof with the A4Tech HS-60 headset microphone interfaced to an Intel Dual Core 2.0 GHz CPU. The software used to record and edit the speech file is wave-pad; ver. 3.05.The recognition experiment is presented in this thesis to obtain comparable recognition results for the speaker-independent recognition of connected sequences of Bangla digit. As the research results has proved that the words accuracy is average 98%.
목차
1. Introduction
2. Research Design
3. Research Methodology
3.1. Description of the Speakers
3.2. Digits Pronunciations
3.3. Data Collection
3.4. Preparation of the Database for the Recorded Pronunciations
3.5. Raw Training and the Preparation of the Testing Data
4. Finding and Results
5. Conclusion
5.1. Recommendation for Further Research
Acknowledgments
References