Singing Songs Identification And Retrieval Technology

In this information for rapid updating of the times, the computer network and multimedia technology has been booming, with the increase in Internet users, more and more people begin to download from the Internet music, with the rock, pop and a series of emerge in an endless stream of song types, making it to retrieve the songs of the increasingly high demand. But the existing search is only limited by the song name, author and text based information browsing, it has been unable to meet the needs of users. A content based music search mode emerge as the times require, humming retrieval refers to the user through the sing sing the Song Melody Lyric or as input to retrieve data from the database, retrieves the corresponding melody music retrieval method; compared with the traditional text based music retrieval, query by humming to provide users with more convenient and friendly inquiry way, not only can text based music retrieval as a useful supplement, but also bring a better user experience.Query by humming relates to many aspects, including database design, extraction of feature parameters, song melody similarity matching problem etc.. In this paper, on the basis of current research achievements, the key problems are studied.Music melody feature extraction including pitch extraction and note segmentation two aspects in pitch extraction, I used a method based on the improved auto-correlation function method, improve computation speed and pitch extraction accuracy. In the note segmentation is adopted based on the energy change and fundamental change note segmentation technology, solves the boundary information fusion to note segmentation difficulties.Humming retrieval algorithm, aiming at the core algorithm has done an in-depth study and on this basis to do some improvements, after verification, the improved algorithm can improve the accuracy and robustness of the system.