The Design and Implementation of Humming Retrieval System Based on Melody

摘要

Music retrieval technology based on humming is a new technology which has a wide application prospect in recent years. The goal of this paper is to construct a complete music retrieval system based on humming. Based on the research of humming music retrieval system, the paper focuses on two major aspects. The first aspect is the processing of humming audio and the extraction of melody features. The second aspect is melody matching algorithm. These two aspects are the core part of humming retrieval system, and also the focus of this paper. For the first aspect of the study, this paper first of the audio signal input from the microphone pretreatment, pretreatment process, including noise removal, pre emphasis and windowing. We adopt the method of setting the threshold value of the volume and the combination of the spectral subtraction, which can be realized by a high pass filter. We adopt the existing methods to implement the audio preprocessing process. As for melody feature extraction, we have done a lot of research work, carried out a large number of simulation experiments, and studied the existing Melody Extraction algorithms. At last, we choose the classical ACF auto correlation method to improve, and realize our own melody feature extraction process. At the same time, in order to build an available music feature library, we also studied the MIDI music melody extraction technology, combine the advantages of several commonly used algorithms, can extract a MIDI melody feature module, and generate the available library file.