Music Genre Classification Using Locality Preserving Non-Negative Tensor Factorization and Sparse Representations

A robust music genre classification framework is proposed that combines the rich, psycho-physiologically grounded properties of auditory cortical representations of music recordings and the power of sparse representation-based classifiers. A novel multilinear subspace analysis method that incorporates the underlying geometrical structure of the cortical representations space into non-negative tensor factorization is proposed for dimensionality reduction compatible to the working principle of sparse representationbased classification. The proposed method is referred to as Locality Preserving Non-Negative Tensor Factorization (LPNTF). Dimensionality reduction is shown to play a crucial role within the classification framework under study. Music genre classification accuracy of 92.4% and 94.38% on the GTZAN and the ISMIR2004 Genre datasets is reported, respectively. Both accuracies outperform any accuracy ever reported for state of the art music genre classification algorithms applied to the aforementioned datasets.