The Design of Audio Fingerprint Identification System
|School||North China University of|
|Course||Signal and Information Processing|
|Keywords||Query By Humming Note segmentation Pitch extraction DTW Linearscaling melody|
It is difficult to use the traditional method of text-based music retrieval to query music when you have just remember a part of the melody of the music and forget the music name, the author and other information. With the development of audio fingerprinting technology, it is regarded as a simple efficient way of music retrieval that if we take melody as input fingerprinting to query the music from the music database. We can retrieve the songs only by the tune of a song or a fragment of it. In a word, the technology of music retrieval by humming is a technology of content-based music in nature.Melody features extraction, note segmentation and approximate melody match is the core of the technologies on music retrieval by humming. The melody features includes the pitch and the duration of notes. For this purpose, we’ve done the following researches:(1) Introduce the research of query by humming, development situation and the basic knowledge of music.(2) According to the mechanism of speech signal, a super resolution pitch detection algorithm which combined Linear Predictive Coding (LPC) with Average Magnitude Difference Function (AMDF) was proposed, The proposed algorithm conquers half frequency errors, and has good accuracy and robustness under the condition of high SNR.(3) In order to improve the retrieval performance of using melody contour and rhythm in Query By Humming, this paper proposed a new algorithm using pitch and energy for note segmentation, the accuracy rate of this algorithm is88.75%in quiet laboratory environment.(4) This paper proposed a new fuzzy matching algorithm using EMD, DTW and LS for Melody matching, it can tolerate the wrongs of notes increase, notes delete and notes replace, achieved a good balance between accuracy and speed.(5) Constructs a humming retrieval system using the above thesis, experimental results and analysis show the system is effective and robust.