Detecting key features in popular music: case study – voice detection

Detecting distinct features in modern pop music is an important problem that can have significant applications in areas such as multimedia entertainment. They can be used, for example, to give a visually coherent representation of the sound. The work developed for this project is meant to be used in the context of a multimedia, multi-touch game where the user has to perform simple tasks at the rhythm of the music. The ultimate goal of this project is to make the creation of visual content automatic using as input, features extracted from the music sound. In this work special focus will be given to the detection of voice segments inside music songs. The solution presented extracts the MFCC’s of the sound and uses a Hidden Markov Model to infer if the sound has voice or not. Using this method some results are presented with the best parameters to extract this feature. Author