Instrument Classification Using Different Machine Learning and Deep Learning Methods

Abstract

Instruments are categorized into the 5 groups in the Sachs-Hornbostel system: idiophones, membranophones, aerophones, chordophones, and electrophones. It might be easy to tell the Sachs-Hornbostel group that an instrument belongs to. However, distinguishing single instrument sound can be hard in monophonic or polyphonic music pieces and it is an important subject for musicians. Using computer science models can help musicians to analyze songs easily and fasten the speed of finding the instrument that are wanted by music producers or composers. This work aims to compare different models on particular instruments (monophonic sound) recognition which is an important problem in the field of music information retrieval. Jupyter Notebook is included for easy reproducibility. Among the six models chosen in this research: k-nearest neighbors(kNN), Support Vector Machines(SVM), Gaussian Mixture Modeling(GMM), Artificial Neural Networks(ANN), Convolutional Neural Networks(CNN) and Recurrent Neural Networks(RNN), CNN is the most accurate model and SVM is the fastest model while CNN has the prospect of being improved because it can be adjusted manually.