Subject: Re: simulated reverberation From: Bradley Wood Libbey <gt1556a(at)PRISM.GATECH.EDU> Date: Wed, 27 Feb 2002 14:35:10 -0500
Dear John, It is difficult to obtain and silence "real" rooms for testing, and I have used simulated room impulse responses. Still, I find myself always referencing back to the real room where a subject can sit and listen to real reverberation. I feel the real room is still the best. But simulations offer much more flexibility in testing and I've found the image method, Allen (1979), to give approximately equivalent speech intelligibility results to a real room (no additive noise). I simplified with an omnidirectional source, did not use frequency specific absorption, nor did I include the digitization correction as proposed by Peterson (1986). There is a slightly artificial quality to the speech, but not too bad, I've considered adding a bit of randomization to the impulses to try to get rid of this artificial sound, but haven't tryed it yet. The frequency specific absorption might also help. Still, this may be more accurate than your test requires. Depending on what you are interested in it may be acceptable to use exponentially decaying noise for the impulse response (or at least beyond the first 30 ms, depending on your room's dimensions). I believe Pierce's book on the fundamentals of acoustics has echo density as a function of time. It has even been suggested that weighted masking noise simulates reverberation for speech, see Gelfand (1979), but this is likely an oversimplification. If speech is your interest the idea of noise replacing reverberation is further addressed in Nabelek's (1989) work. As for the moving air, this will have a minimal contribution provided the air velocity is much less than the speed of sound. Perhaps you have heard that the motion of the speaker or listener will drastically change the impulse response, which is true. As will temperature and humidity changes (this is problem for signal processes trying to cancel reverberation.) In the above cases I would propose that the impulse response in nonstationary or time dependent, but still linear. No matter what method, it will never be perfect. In my mind the question is, what aspects need to be included given what the test is trying to determine [intelligibility, quality, localization, music, environmental sounds?] The following may or may not need to be considered. -reverberation time -room dimensions -listener position, velocity, head rotation, clothing -absorption coefficients four walls, floor, ceiling. (freq. dependent?) -source position (inside/outside critical distance) -source directivity -objects within the room -temperature humidity -time dependence Hope this will stimulate some thoughts, Brad Libbey Georgia Institute of Technology Allen, J. B., and Berkley (1979). "Image Method for Efficiently Simulating Small-Room Acoustics." J. Acoust Soc. Am. 65: 943-950. Gelfand, Stanley A. and Shlomo Silman. (1979). "Effects of small room reverberation upon the recognition of some consonant features." J. Acoust. Soc. Am., 66: 22-29. Nabelek, Anna K. , Tomasz R. Letowski, and Frances M. Tucker. (1989). "Reverberant overlap- and self-masking in consonant identification." J. Acoust. Soc. Am., 86: 1259-1265. Peterson, P. M. (1986). "Simulating the Response of Multiple Microphones to a Single Acoustic Source in a Reverberant Room." J. Acoust Soc. Am. 80: 1527-1529. Pierce, Allen D. (1991). Acoustics, and Introduction to Its Physical Principles and Applications, Woodbury, NY: Acoustical Soc. of America.