A Bayesian approach to time-frequency based blind source separation
Cédric Févotte, Simon Godsill · 2006
Contained herein is a brief description of three audio files that accompany the paper "The Binaural Display of Clouds of Points Sources" (Adams, N and Wakefield, G. 2005). The three audio files contain examples of the clouds of 'popcorn pops' that were synthesized for the paper. The examples motivate the display of clouds of point sources, as well as compare the different auralization methods presented in the paper. All three audio files are constructed by first generating a 20 ms pink noise burst (bandpass filtered to 150 Hz to 1.5 kHz). The monaural signal is then filtered with either a binaural pair of measured HRTFs or a low-order approximation thereof. The final audio files are then generated by triggering the binaural noise burst according to a Poisson process and scaling each burst with a random 0-18 dB amplitude factor, yielding a cloud of click processes. The names of the three WAV files are, Points/spl I.bar/vs/spl I.bar/Clouds.wav; Compare/spl I.bar/Clouds/spl I.bar/Same.wav; and Compare/spl I.bar/Clouds/spl I.bar/Diff.wav.