DEVICE AND A METHOD FOR EXTRACTING A TARGET SOUND FROM A MIXTURE SOUND, CAPABLE OF PREVENTING AN INTERFERENCE SOUND FROM BEING REMOVED WHILE A TARGET SOUND IS DISTORTED

摘要

PURPOSE: A device and a method for extracting a target sound are provided to accurately extract a target sound from a mixture sound.;CONSTITUTION: A modeling system(101) extracts a base matrix of training noise from training noise. A sound decomposition unit(102) decomposes a received mixture sound into a target sound and an interference sound using the base matrix. The interference sound is modeled by the base matrix. The modeling unit converts the training noise into a time-frequency domain. The modeling unit performs non-negative matrix factorization of the converted training noise.;COPYRIGHT KIPO 2011