Output details

A multistage approach to blind separation of convolutive speech mixtures

The 'Cocktail party problem' is a long-standing scientific challenge, as it is difficult to automatically extract/separate a conversation from multiple conversations. This paper presents a new method based on the integration of convolutive blind source separation and ideal binary masking, together with cepstral smoothing. This algorithm has been used in the Binaural Hearing Aid systems developed jointly by Universidad del Quindio and Georgia Institute of Technology; the Close-Talk System developed jointly by The Quartermaster Equipment Research Institute and Tsinghua University; Speech Enhancement System developed by University of Texas and the Auditory Prostheses system developed by Katholieke University.