Output details

A Multiplicative Algorithm for Convolutive Non-Negative Matrix Factorization Based on Squared Euclidean Distance

This paper explores non-negative matrix factorisation (NMF), which has been widely used in signal/image processing and pattern recognition. NMF is limited for monitoring speech and audio. The proposed new framework generalises NMF into a convolutive model and has been employed in transcription and coding systems because of its superior performance for audio onset detection and pattern separation. The proposed algorithm has been employed in audio separation, recognition, transcription and coding systems and has been used by the EURECOM and Technische Universitat Munchen achieving state-of-the-art results in Munich 2011 CHiME Challenge on speech enhancement and recognition in multi-source and noisy environments.