Sound textures are a specific category of sounds that originates from the superimposition of a large number of similar acoustic events. Typical examples of sound textures are the sound of the rain or the fire cracking. Due to their intrinsic properties, sound textures offer the opportunity to apply computational modeling methods and assess how the auditory system extracts abstract representations. Consistent evidence exists that the auditory system processes sound textures by two main modes of representation. First, local spectral features are extracted. As the amount of information increases, statistics are averaged across time and local details are no longer retained, determining information abstraction. The aim of our study was to assess whether the local feature analysis and the statistical averaging process are prone to changes due to altered sensory experience or whether they entirely rely on mechanical and deterministic properties of the auditory system. To address this issue, we used the model of permanent visual deprivation. Previous studies have documented that blind individuals outperform sighted controls in a number of auditory tasks, such as: short-term memory tasks, ultra-fast speech comprehension, discrimination of speech in noisy environments. However, which type of auditory computations are affected by an altered sensory experience is unknown. . To the best of our knowledge, this study represents the first attempt to answer to this question by using a computational synthesis approach to assess experience dependent plasticity of the auditory system. Two groups of participants (matched for gender, age, and education) were behaviorally tested. A group of 18 sighted individuals (SC) and a group of 18 blind individuals. Blind participants were divided in two clusters according to blindness onset: nine congenitally blinds (CB) and nine late onset blinds (LB). Using a synthesis algorithm (McDermott and Simoncelli, 2011) we synthesized four exemplars per each of thirty-six sound textures. Two experiments were performed: (i) in the exemplar discrimination experiment they had to discriminate different exemplars of the same texture; (ii) in the texture discrimination experiment they were asked to discriminate excerpts coming from different textures. Stimuli of different durations were presented to dissociate the efficiency of local feature analysis and of statistical averaging processes in each of the experiments. SC performed as expected: a better performance for stimuli of short durations in the exemplar discrimination task and for long durations in the texture discrimination task was found. No significant difference was observed between CB and SC in any of the experiments. However, LB performed significantly worse than both CB and SC, selectively in the exemplar discrimination task. The results of the present study suggest that (1) local feature analysis and the statistical averaging process seem resilient to the lack of vision since birth, (2) late onset visual deprivation instead, might have a selective impact on the ability to retain local features details of sound information. We speculate that late onset blind individuals might learn to favor statistical averaging process and thus, reduce the local feature analysis to allow an early sound objects identification.
|Titolo:||Experience dependent plasticity of auditory statistics: a computational approach.|
|Data di pubblicazione:||2019|
|Appare nelle tipologie:||4.3 Poster|