Emerging Techniques for Sound Classification of Musical Instruments: A Review

Abstract

Musical Information Retrieval (MIR) is an emerging field. It is extremely required to know what instruments are used in an audio file. Musical instruments sound classification has become very popular in the research field. It is always a challenging aspect for researchers because the sound produced by each instrument has different patterns, waveforms, frequencies, and pitch. In this article, various machine learning and deep learning techniques are presented for the identification of monophonic and polyphonic musical instruments playing in audio files. Some of the emerging machine and deep learning techniques are Support Vector Machine, k- Nearest Neighbour, MFCC, Neural Network, Convolutional Neural Network, Denoising Autoencoder, Wavelet Transform, Random Forest, Naive Bayes, Residual Neural Network, GoogleNet, Discrete Wavelet Transform. These approaches proved to be good for instrument classification. There is scope to carry out further research in the identification and classification of polyphonic musical instruments playing together in the audio file and also there is a necessity for detection of overlapping instruments. This paper contributes different methods invented to perform sound classification of musical instruments and also gives a brief idea about scope of the research.