DEVELOPMENT OF ISOLATED SPEECH RECOGNITION SYSTEM FOR BANGLA WORDS
暂无分享,去创建一个
This research devoted to the development of Speech Recognition System in Bengali language that works with speaker independent, isolated and subword-unit-based approaches. In our work, the original Bangla speech words were recorded and stored as RIFF (.wav) file. Then these words were classified into three different groups according to the number of syllables of the speech words and these grouping speech signals were converted to digital form, in order to extract features. The features were extracted by the method of Mel Frequency Cepstrum Coefficient (MFCC) analysis. The recognition system includes direct Euclidean distance measurement technique. The test database contained 600 distinct Bangla speech words and each word was recorded from six different speakers. The development software is written in Turbo C and common feature of today’s software have been included. The development system achieved recognition rate at about 96% for single speaker and 84.28% for multiple speakers. Keywords: MFCC; Syllable-based grouping; Speaker independent; End-point detection; Euclidian distance. DOI: http://dx.doi.org/10.3329/diujst.v6i1.9331 DIUJST 2011; 6(1): 30-35
[1] Lalit R. Bahl,et al. Design of a linguistic statistical decoder for the recognition of continuous speech , 1975, IEEE Trans. Inf. Theory.
[2] Mumit Khan,et al. Isolated and continuous bangla speech recognition: implementation, performance and application perspective , 2007 .
[3] S. Gokul. Multimedia Magic , 2003 .
[4] Jean-Claude Junqua,et al. Robustness in Automatic Speech Recognition: Fundamentals and Applications , 1995 .