Convolutive Blind Source Separation of Speech Signals in the Low Frequency Bands

Jafari MG, Plumbley MD. Convolutive Blind Source Separation of Speech Signals in the Low Frequency Bands. In: AES Convention 123. Audio Engineering Society; 2007. Paper 7312. Available from: https://aes.org/publications/elibrary-page/?id=14369

@inproceedings{Jafari2007_14369,
  author = {Jafari, Maria G. and Plumbley, Mark D.},
  title = {{Convolutive Blind Source Separation of Speech Signals in the Low Frequency Bands}},
  booktitle = {AES Convention 123},
  note = {Paper 7312},
  year = {2007},
  month = oct,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=14369}
}

TY  - CPAPER
TI  - Convolutive Blind Source Separation of Speech Signals in the Low Frequency Bands
AU  - Jafari, Maria G.
AU  - Plumbley, Mark D.
T2  - AES Convention 123
M1  - Paper 7312
PY  - 2007
DA  - 2007/10/06
UR  - https://aes.org/publications/elibrary-page/?id=14369
PB  - Audio Engineering Society
LA  - en
AB  - Sub-band methods are often used to address the problem of convolutive blind speech separation, as they offer the computational advantage of approximating convolutions by multiplications. The computational load, however, often remains quite high, because separation is performed on several sub-bands. In this paper, we exploit the well known fact that the high frequency content of speech signals typically conveys little information, since most of the speech power is found in frequencies up to 4kHz, and consider separation only in frequency bands below a certain threshold. We investigate the effect of changing the threshold, and find that separation performed only in the low frequencies can lead to the recovered signals being similar in quality to those extracted from all frequencies.
ER  -