Exploring Nonnegative Matrix Factorization for Audio Classification: Application to Speaker Recognition

In this paper, we test the use of Nonnegative Matrix Fac-torization (NMF) for feature extraction in the context of audio classification. NMF calculates a decomposition of the spectrogram into nonnegative factors and has been successfully applied to audio source separation. Thus, it has the potential to be robust to noise disturbances when used for feature calculation. We then introduce two fea-ture sets directly derived from the NMF decomposition. Experiments performed on an 8-class speaker recognition task with Support Vector Machines show that the pro-posed representations convey complementary information to the baseline MFCC features. Indeed, the use of only the NMF-based descriptors lead to similar results as the refer-ence features, and the combination of these representations yields a significant improvement of the obtained accuracy.