Automatic Instrument Detection in Audio Signals by Comparing Deep Learning Network Architectures

Abstract

There are many different types of signals, including electrical, video, and audio signals, each with distinct significance and applications. The proposed aim of the study is to develop and evaluate effective methods for automatic instrument detection in audio signals. The ultimate goal is to improve applications such as database indexing, song annotation, and tools for musicians and music producers.In this task, we solved using Mel-frequency cepstral coefficients (MFCC) and several neural network architectures, namely, artificial neural network (ANN), convolutional neural network (CNN), MLNet, and the committee of voting classifiers compared in this work. The results show that the proposed methods achieved the best classification quality using an extensive model, which is the committee of voting classifiers.