A Review on Speech-Music Discrimination Methods
A. Masoumeh Velayatipour, B. Mohammad Mosleh · 2014
Automatic speech-music separation has been become as one of attractive and high applicable topics, in recent years. Classifying the sound into categories such as speech or music is known as an important and basic requirement in multimedia document retrieval systems. In today world, the range of audio sources is very wide and it can be said that the audio data range from broadcast news and television to diverse audio recorded on magnetic tapes or even the audios stored in the web. This diversity has encountered the music-speech separation with some challenges. Therefore, separating the speech from music, before any additional processing, is the first step which is performed on the audio data. In speech-music separation field; some different researches have been done to extracting efficient features from audio signals or proper classifying the audio signals into two categories, speech and audio. In this paper we investigate different methods presented in this area and compare their results. The results show that the combinational systems based on fuzzy-evoluti onal rules have better discrimination accuracy, compared with other methods.