Research And Implementation Of Content-Based Music Retrieval System

Content-based music retrieval is a way to retrieve music from the content contained in the music itself by analyzing the music and audio features music signal. Humming-style music retrieval methods for user requirements low skills, easy to use and versatile, so it is called the most important way of the current music retrieval.Humming Music Retrieval include feature extraction and matching melody two modules. In this paper, we research and analysis the feature extraction and melodic melody matching algorithm and choose a different algorithm based on the improved performance of the algorithm developed a prototype system to verify the effectiveness and performance of the algorithm. The main works include:1) Establish the music database and extract the theme of every music.2) Research the humming audio feature extraction technologies, including audio signal processing, genetic testing, notes cut grading work. The extracted audio feature can be saved as a sequence of audio frames and sequence of notes.3) Select different melody matching algorithms and try to improve the algorithm, making the retrieval speed and accuracy of the matching algorithm has good performance.4) Design and implement a prototype system for verification effectiveness and accuracy of the matching algorithm.