Spectral Sub-Band Filter Dependent Windowing Music Genre Classification Machine Learning Dataset
Description
This dataset contains the generated features and targets used for training music genre classification systems in [1]. The data contains mean and standard deviation summaries of the sub-band filter-dependent windowing features extracted from the fault-filtered version of the GTZAN audio genre dataset [2]. Fault filtering specifications are described in [1]. Data is in CSV format; feature names contain the following suffixes: f (1-10), mean or std. The 'f' suffix specifies the sub-band filter index while 'mean' or 'std' is the statistical summary. Please refer to the source [1] for the extraction specification details. References: [1] F. Prezja, “Developing and testing sub-band spectral features in music genre and music mood machine learning,” Master Thesis, University of Jyväskylä, Jyväskylä, November. 2018. [Online]. Available: https://jyx.jyu.fi/bitstream/handle/123456789/60963/1/URN%3ANBN%3Afi%3Ajyu-201901081104.pdf [2] G. Tzanetakis and P. Cook, “Musical genre classification of audio signals,” IEEE Trans. Speech Audio Process., vol. 10, no. 5, pp. 293–302, Jul. 2002.
Files
Steps to reproduce
All feature extraction specification details can be found in [1].