Abstracts Submitted: 250
Number of Users: 297
View Abstracts Submitted
Back to home Page
At present Automatic Speaker Recognition system is a very important issue due to its diverse applications. Hence, it becomes absolutely necessary to obtain models that take into consideration the speaking style of a person, vocal tract information, timbral qualities of his voice and other congenital information regarding his voice. The study of Bengali speech recognition and speaker identification is scarce in the literature. Hence the need arises for involving Bengali subjects in modelling our speaker identification engine. Being subdivided into many dialects, the language Bengali has different phraseologies corresponding to particular regions. Thus we selected subjects from different regions of Bengal for our research. Different dialects are expected to reflect much contrasting acoustic features. A pool of 10 participants have been selected for this study who are comfortable with different dialects, and a total of 8 Bengali speech database was recorded from each of them. The text database each of about 3 min in length have been selected from famous Bengali writings and are supposed to convey a meaning even in the speech emotional perspective. In this work, we have extracted some acoustic features of speech like vowel onset points, pitch contour, formant frequencies, timbral qualities for the purpose of segmentation of Bengali speech. Classification and identification of the corresponding speakers also form another aspect of this study. The spectral features which form the basis of this model include spectral envelope, shimmer, jitter, inharmonicity, tristimulus and spectral centroid. Temporal features such as attack, steady-state, decay and vowel onset points are also analysed and incorporated in this study. We obtain the vocal tract information by means of Formant frequencies. Accumulation of energy around a particular frequency (Formants) which corresponds to resonance in vocal tract differs from person to person. A Confusion Matrix corresponding to our obtained results clearly specifies the dissimilarities of subjects from different dialects. The Results obtained from this study gives a good recognition rate for Bengali Speakers.
© Copyright 2017 All Rights Reserved