Quantifying the just noticeable difference of reverberation time with band-limited noise centered around 1000 Hz using a transformed up-down adaptive method

Authors

Matthew G. Blevins, University of Nebraska-LincolnFollow
Adam T. Buck, University of Nebraska-LincolnFollow
Zhao Peng, University of Nebraska-LincolnFollow
Lily M. Wang, University of Nebraska - LincolnFollow

Date of this Version

6-9-2013

Citation

Blevins, M., A.T. Buck, Z. Peng, and L. M. Wang. (2013) “Quantifying the just noticeable difference of reverberation time with band-limited noise centered around 1000 Hz using a transformed up-down adaptive method.” Proceedings of the International Symposium on Room Acoustics; Toronto, Canada; June 9-11, 2013.

Comments

Copyright (c) 2013 Matthew G. Blevins, Adam T. Buck, Zhao Peng, & Lily M. Wang.

Abstract

This study seeks to quantify the just noticeable difference (JND) of reverberation time (RT) using band-limited noise. ISO 3382-1 lists the JND of reverberation metrics at 5% based on work by Seraphim (1958). However, others have found the JND of RT to be higher from 6% to 39%. Many of these studies utilized band-limited stimuli, e.g. speech, music motifs and bandlimited noise. A previous study by the authors conducted on 30 subjects using white noise demonstrated a JND of RT at 22%. To further verify these results and investigate potential upward frequency masking, the present study was conducted following the same methodology but using octave-band limited noise centered at 1000 Hz instead of white noise. Binaural room impulse responses (BRIR) were created from the Elmia concert hall in ODEON by uniformly varying absorption coefficients across all surfaces and frequencies to achieve the desired RTs. The desired RTs varied around three reference values (1, 2, and 3 seconds), with eight samples approaching the reference RT from below and another eight approaching from above, at 4% intervals of the reference RT. Auralizations of the BRIRs and 500 ms band-limited noise were randomly presented in a computer-based testing program using a three-interval one-up twodown forced choice method, while interleaving six staircase sequences (3 reference RT X 2 downward vs. upward approaching direction). Subjects were individually tested in a sound attenuated booth using headphones with flat frequency response. Results are presented and compared against those previously obtained using white noise.