Blind Vector Deconvolution: Convolutive Mixture Models in Short-Time Fourier Transform Domain

摘要: For short-time Fourier Transform (STFT) domain ICA, dealing with reverberant sounds is a significant issue. It often invites dilemma on STFT frame length: frames shorter than reverberation time (short frames) generate incomplete instantaneous mixtures, while too long may disturb the separation. To improve separation of such sounds, authors propose new framework which accounts for short frames. In this framework, convolutive mixtures are transformed to mixtures. separating an approach applying another presented so as treat them mixtures. The experimentally confirmed that outperforms conventional ICA.