作者: Jack Xin , Wenye Ma , Meng Yu , Stanley J. Osher
DOI:
关键词: Convexity 、 Artificial intelligence 、 Regularization (mathematics) 、 Source separation 、 Convex optimization 、 Pattern recognition 、 Computer science 、 Bregman method 、 Extraction (chemistry)
摘要: A fast speech extraction (FSE) method is presented using convex optimization made possible by pause detection of the sources. Sparse unmixing filters are sought l1 regularization and split Bregman method. subdivided developed for efficiently estimating long reverberations in real room recordings. The based on a binary mask source separation FSE evaluated found to outperform existing blind approaches both synthetic recorded data terms overall computational speed quality. Index Terms: convexity, sparse filters, method, extraction.