The Design And Implementation Of Music Retrieval System Based On Rhythm Information

With the rapid development of network technology and multimedia technology, Various forms of resources are showing a vigorous growth trend. In the various multimedia resources. Music, because of its special form and application, is favored by researchers. In order to efficiently access our favorite music resources, convenient tools for music browsing and query are needed. Content-based music retrieval technology is based on the content of the music itself to complete the search information. It achieves a friendly human-computer interaction better than other ways,and the use of fuzzy matching is more adapted to the music data changes.when the user does not remember the music text information, by humming or providing a sample of the music, the content-based music retrieval technology can better retrieval music, which cannot be retrieved by text-based technology.Rhythm is an important part of music, which includes the information such as pitch length, pitch and main melody of music. It is of great value to extract and match the prosody of music. In this paper, the research is according to the current trend of music retrieval and demand,The key technology of retrieving music based on the content of music lies in two parts, the music feature extraction and feature matching, which are the main content of this research.This paper first introduces the basic knowledge of music, according to the knowledge of music, the detailed design scheme of music retrieval system is proposed, and the feature extraction algorithm and matching algorithm are analyzed. The main contents of this paper include the following:(1) Feature extractionThe feature extraction of music data is an important research content of this topic. GFCC features, MFCC features, rhythm features and main melodic features of the data are extracted, in addition, under the procession of Principle component analysis, fusing the GFCC and rhythm features to obtain new musical feature data.(2) The fusion methodAfter learning the main melodic features, MFCC, GFCC and rhythm features. This paper shows a new fusion method about fusing GFCC feature and rhythm feature. Through the experimental comprasion, the proposed method could raise the retrieval accuracy of music.(3) Music retrievalIt is important for this paper to study music retrieval algorithm.Nowadays, there are so many mature retrieval algorithms, Dynamic Time Working method has the ability to overcome the decrease of accuracy due to the changing of rate, and that is the reason this paper chooses Dynamic time warping algorithm to retrieve music. Above all, K-Nearest Neighbor(KNN) algorithm is also tried to achieve the classification of music retrieval.(4) The design of systemAfter researching and improving the algorithm, a system is designed to apply these algorithms. 1200 pieces of music are downloaded from internet including different language and styles.