A Real-Time Music Retrival System Based On Melodic Feature

As a product of human emotional expression,music has become a part of people's lives.At present,although the music retrieval technology has been widely used,but the traditional music retrieval technology still has some defects and shortcomings,for example,it is not based on the characteristics of the music itself,and needs to spend a lot of manpower and time to mark each song,more importantly,in the case of which the user only remember the melody,the traditional retrieval system to this helpless.Therefore,in view of the shortcomings and shortcomings of the traditional music retrieval technology,this paper presents a humming retrieval system based on the compound music database,which uses the melody characteristics of the music itself to identify the music,is a more convenient and more efficient music Retrieve the way.Melody is the soul of music,is the most widely used and most representative feature in the identification of music.Based on the existing achievements and techniques,this paper realizes a humming retrieval system based on the melody features of music,and carries on the research in the following aspects:(1)A retrieval system based on polyphonic music database is proposed and implemented.The database is MIDI database in the traditional humming retrieval system based on the melody.The MIDI does not contain vocals,and it is not realistic for every song to make MIDI.Therefore,the humming retrieval system based on the polyphonic music database has expanded the application scope,and has more practical value.(2)The algorithms of fundamental frequency extraction of humming and the vocal polyphonic music are studied respectively.An improved cepstrum method humming pitch extraction is proposed,and the comparison of two kinds of polyphonic music pitch extraction algorithm is presented,and the MELODIA algorithm is used to extract pitch from polyphonic music.(3)A new matching engine design scheme combining LS and DTW algorithm is proposed.The matching algorithm is the core of the humming retrieval system.The change of humming speed,the quality of humming and tone errors are important factors to consider about the matching algorithm.The minimum distance principle is used to combine LS with DTW,which is less retrieval time and higher accuracy of retrieval.(4)The humming retrieval system proposed in this paper is realized based on the C/S model.font-end of the mobile is based on the Android platform,which contains recording and VAD module.The back-end of the server is based on the Linux platform,which contains the extraction of the fundamental frequency,the management of the database,the matching of the fundamental frequency sequences and so on.A large number of system tests have been carried out,and the results show that the system has good performance.