期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2015
卷号：6
期号：4
页码：3447-3451
出版社：TechScience Publications
摘要：The proposed work deals with the practise of multi scale analysis of a vocal music signal. The wavelet domain can be introduced to classify an exclusive singer from a database containing different singers. As it is well known that the wavelet transform of a signal at low scale value corresponds to the approximate of the signal in the original form but more details of the signal can be achieved with the analysis obtained at higher scale values of the wavelet transform. Keeping in this property of the wavelet transforms, a novel method is proposed that fuses the information obtained from the different scale wavelet transforms of an audio signal, and then use this method to match different singers in a database with a particular singer. Results show that the proposed method achieves more accuracy as compared to direct analysis or even a single scale analysis of a signal.
