Fast Multichannel Nonnegative Matrix Factorization With Directivity-Aware Jointly-Diagonalizable Spatial Covariance Matrices for Blind Source Separation.
Kouhei SekiguchiYoshiaki BandoAditya Arie NugrahaKazuyoshi YoshiiTatsuya KawaharaPublished in: IEEE ACM Trans. Audio Speech Lang. Process. (2020)
Keyphrases
- nonnegative matrix factorization
- blind source separation
- negative matrix factorization
- covariance matrices
- covariance matrix
- principal component analysis
- maximum likelihood
- independent component analysis
- matrix factorization
- distance measure
- sparse representation
- gaussian distribution
- vector space
- document clustering
- non stationary
- least squares
- spectral clustering
- pattern recognition
- data representation
- feature vectors
- multiscale
- missing data
- data analysis