Real and Virtual Instruments in Machine Learning –Training and Comparison of Classification Results

The continuous growth of the computing power of processors, as well as the fact that computational clusters can be created from combined machines, allows for increasing the complexity of algorithms that can be trained. The process, however, requires expanding the basis of the training sets. One of the main obstacles in music classification is the lack of high-quality, real-life recording database for every instrument with a variety of musical articulation. This study aims not only to expand databases with samples prepared by using VSTs but also to verify quality of real-instrument signal classification, compared to signals created digitally. Also, the possibility of training an algorithm with real and additional synthetic data mixed is explored. In the paper, machine learning algorithms are described first. A way to generate an expanded training set and the classification process preparation is presented afterward. Then, the results are compared, and future work is outlined in conclusions.