The Research And Application Of The Note Recognition Algorithm Based On Nonlinear Characteristic

Notes recognition is the important research content in the field of musical signals processing. Besides, notes recognition technology has important application value for creation, management and index of music database. This thesis has built the recognition method from the point of nonlinear characteristics.This thesis investigates the relevant music recognition methods and shortages of frequency analysis are summarized. Based on traditional methods, the process of characteristic parameter and classifier has been build. In choosing the characteristic parameter, nonlinear characteristic is obtained through Reconstruction of Phase Space, which has been used in the note recognition experiments. The results shows that the way to obtain the characteristic through fixed parameter would make information be lost for some kinds of note, which will lower the recognition accuracy. In order to get the better characteristic, adaptive signal decomposition has been introduced in the preprocessing which will decompose the original signal into several characteristic sub-signals. Meanwhile, in order to improve the efficiency and reduce the recognition time, principal component which obtained from the nonlinear matrix by PCA will used in structuring classifier and experiments.The results of relevant experiments show that the music notes recognition based on nonlinear characteristics can distinguish the type of notes accurately, which offers a feasible method to the relevant areas such as music database.