Audio Processing Technology Based on Sift in Music Teaching

Abstract

Audio processing technology plays an important role in music teaching. Music teaching materials come from a large number of audio materials, and voice / music discrimination is an important step in audio processing and analysis, such as efficient audio coding, audio retrieval, automatic speech recognition, etc. This paper presents a speech/music segmentation and classification method based on Scale-invariant feature transform (SIFT). Firstly, the change points of audio are detected according to the difference of mean square energy between adjacent frames to realize segmentation. Then, eight dimensional features such as low band energy variance ratio, cepstrum energy modulation and entropy modulation are extracted and classified by artificial neural network. Experimental results show that the proposed algorithm and features have high segmentation accuracy and classification accuracy. The audio processing technology has a certain reference value for the classification of music teaching materials, and has a certain auxiliary role for music teaching.