Automatic music mood classification via Low-Rank Representation
Yannis Panagakis, Constantine L. Kotropoulos · European Signal Processing Conference · 2011
The problem of automatic music mood classification is addressed by resorting to low-rank representation of slow auditory spectro-temporal modulations. Recently, it has been shown that if each data class is linearly spanned by a subspace of unknown dimensions and the data are noiseless, the lowest-rank representation (LRR) of a set of test vector samples with respect to a set of training vector samples has the nature of being both dense for within-class affinities and almost zero for between-class affinities. Consequently, the LRR exactly reveals the classification of the data, resulting into the so-called Low-Rank Representation-based Classification (LRRC). The performance of the LRRC is compared against three well-known classifiers, namely the Sparse Representations-based Classifier, Support Vector Machines, and Nearest Neighbor classifiers for music mood classification by conducting experiments on the MTV and the Soundtracks180 datasets. The experimental results validate the effectiveness of the LRRC among the classifiers that is compared to.