MScProject : A Comparative Study of Supervised Machine Learning Techniques for Automated Instrument Classification
In order to reduce the time spent on manual classification of the category of sounds, various Machine Learning techniques can be utilised. In this research paper, we have developed an algorithm to categorise instruments from a dataset into their respective classes, utilising extracted information from descriptors of audio, and by implementing various supervised methods of Machine Learning, namely Naive Bayes (12.59% accuracy), Random Forests (73.35% accuracy), and Support Vector Machine (3.4% accuracy). To conclude the research, we present a few comparisons in terms of performance of these algorithms. The presented information includes relevant factors to be considered regarding the viability of classifying instruments automatically.