Abstract
The perception of reverberation is constrained by environmental statistics
The Journal of the Acoustical Society of America, Vol.139(4 Supplement), pp.2210-2210
04/2016
DOI: 10.1121/1.4950603
Abstract
Human sound recognition is robust to reverberation. We explored the hypothesis that this robustness is rooted in the ability to separate the contributions of a sound’s source from that of reverberation. As the separation of source and filter from their convolution is inherently ill-posed, any such capacity should depend on prior assumptions about the nature of filter and/or source. We measured the distribution of real-world environmental impulse responses (IRs) and tested whether it constrains the ability of listeners to estimate source and filter from reverberant audio. We surveyed volunteers about the spaces they encountered during daily life, and measured IRs at each location. We found that the tails of real-world IRs decay exponentially, with decay rates consistently slower at low frequencies. We then synthesized IRs that were either faithful to, or deviated from, the observed distribution of real-world IRs. We assessed (a) the sense of reverberation conveyed by IRs, (b) discrimination of novel sound sources in reverberation, and (c) discrimination of IRs given only their convolution with sound sources. We found that human listeners can separately estimate the source and filter in reverberant conditions, but are strongly constrained by whether the filter conforms to the naturally occurring distribution.
Details
- Title: Subtitle
- The perception of reverberation is constrained by environmental statistics
- Creators
- James Traer - Brain and Cognit. Sci., MIT, 77 Massachusetts Ave., Cambridge, MA 02139, jtraer@mit.eduJosh H McDermott - Brain and Cognit. Sci., MIT, 77 Massachusetts Ave., Cambridge, MA 02139, jtraer@mit.edu
- Resource Type
- Abstract
- Publication Details
- The Journal of the Acoustical Society of America, Vol.139(4 Supplement), pp.2210-2210
- DOI
- 10.1121/1.4950603
- NLM abbreviation
- J Acoust Soc Am
- ISSN
- 0001-4966
- eISSN
- 1520-8524
- Publisher
- Acoustical Society of America
- Number of pages
- 1
- Language
- English
- Date published
- 04/2016
- Academic Unit
- Psychological and Brain Sciences; Iowa Neuroscience Institute
- Record Identifier
- 9984065470402771
Metrics
23 Record Views