An Automatic Classification Method For Western Musical Instruments Based On Sparse Representation And Deep Neural Network Model

With the rapid development of modern social science and technology,the number of digital music also grows massively.In order to facilitate users' music retrieval,effective and reasonable classification of music is very important.But due to the large base of music,the number of new single songs continues increasing every day.It's impractical to take artificial way to classify the music.Therefore,this paper uses the deep neural network model to extract the characteristics of music to achieve automatic classification.At the same time,due to the sparse feature of the music signal,this paper proposes a new method to combine the sparse features and the deep neural network model to realize the automatic classification of music signals.This paper aims to apply the new method to automatic classification of western musical instruments.This paper first introduces the common characteristics of music signals,including the basic physical characteristics and the psychological characteristics.At the same time it also describes human auditory system.Then it introduces the theory of automatic music classification,including the feature extraction of music,and emphatically introduced the relevant concepts of deep neural network model used in this paper.Then,we study the sparse feature's extraction process of music signal,and introduce a new sparse representation dictionary,which is based on the difference between different musical instruments.In this paper,we compare the results of sparse reconstruction based on the traditional dictionary and the dictionary being constructed in this paper,and prove our new dictionary is better than the traditional dictionary.Finally,this paper studies the automatic classification of western musical instrument based on the deep neural network model.Traditionally,the input of the deep neural network model is the Mel Frequency Cepstrum Coefficient(MFCC)of the audio signal.This paper creatively uses the audio signal's sparse feature as the deep neural network's input.And then,we train the model's parameters to realize the automatic classification of western musical instruments.In this paper,we use "Python" programming language and our final accuracy of automatic classification of western musical instrument based on sparse features and deep neural network model could reach to 82%.