Discovering Convolutive Speech Phones Using Sparseness and Non-negativity

Discovering a representation that allows auditory data to be parsimoniously represented is useful for many machine learning and signal processing tasks. Such a representation can be constructed by Non-negative Matrix Factorisation (NMF), which is a method for finding parts-based representations of non-negative data. Here, we present a convolutive NMF algorithm that includes a sparseness constraint on the activations and has multiplicative updates. In combination with a spectral magnitude transform of speech, this method extracts speech phones that exhibit sparse activation patterns, which we use in a supervised separation scheme for monophonic mixtures.

• Type: Conference
• Year: 2007
• Where: ICA
• Authors: Paul D. O'Grady, Barak A. Pearlmutter
• Keywords: Convolutive Nmf Algorithm, ICA 2007, Method Extracts Speech, Non-negative Matrix Factorisation, Signal Processing