Generating a digital signature for singers to identify their songs

Abstract

A counterfeit is an imitation of the voice of a popular artist, done with the intention of selling or passing it on as a genuine. This imitating of songs from the original artists is being done very smart and smooth, so it becomes impossible to detect it as real or fake. These wrongdoers make an income by selling songs which are imitated disguised as originals. This study proposes a solution for this problem by providing digital signatures for singers that are generated using songs sung by the artists. The songs contain vocal signals surrounded with instrumental music. In order to generate signatures for the voice of the singer, the vocals have to be isolated. This study proposes an isolation technique, which is proved against a prevailing technique. The signature is generated by using the features extracted after voice isolation. The signature of the singer is originated as a Gaussian Mixture Model. The project had been implemented using open source software. The evaluation had been performed through quantitative and qualitative approaches. The outcome of this research had been successful in generating digital signatures for singers. The singers had been identi ed accurately even for those who possess similar voices.