Stochastic resonance can be used as a measuring tool to quantify the ability of the human brain to interpret noise contaminated visual patterns. Here we report the results of a psychophysics experiment which show that the brain can consistently and quantitatively interpret detail in a stationary image obscured with time varying noise and that both the noise intensity and its temporal characteristics strongly determine the perceived image quality. [S0031-9007(97)02344-2]
We describe the results of computer simulations of the dynamical behavior of an autoassociative network with a two-dimensional energy landscape.Such a network can model some aspects of the phenomenon of perceptual bistability in the presence of ambiguous figures. The network can be operated at either zero or nonzero temperatures which represent an internal system noise. Our results show that, under the inAuence of a weak periodic external signal, the network exhibits a maximum in the signal-to-noise ratio at an optimum noise level: the characteristic signature of stochastic resonance.Studies of the perception of ambiguous figures have a long history [1,2]. Perception of these kinds of figures (e.g., the Necker cube [2]; see Fig. 1) is characterized by noisy bistable dynamics, that is, the two different interpretations, elicited by the figure, are alternatively perceived by the observer with a stochastic time course. N umerous experiments have shown that the times between such reversals are approximately gamma distributed [3]. Such distributions are common in biology and can be interpreted in terms of the noise driven motion of a state point which randomly crosses a threshold or surmounts an energy barrier. Recently, there has been a revival of interest in these results in connection with the development of dynamical models of brain function during such reversals in perception [4]. There has also been a growing interest in stochastic resonance (SR) associated with noisy nonlinear systems [5]. This is a dynamical behavior wherein noise may enhance the transmission of information through certain systems, such that a defined signal-to-noise ratio (SNR) achieves a maximum for an optimum value of the noise intensity. SR has been demonstrated in numerous physical experiments [5,6] and, more recently, in a simple sensory neuron [7]. The I IG. 1. Necker cube with its two alternative interpretations. theory of SR was first advanced as a possible explanation of the observed periodic recurrences of the Earth's Ice Ages [8], and, stimulated by an interesting experiment with a bistable ring laser [9], has been the object of numerous recent theoretical studies [10]. The possible importance of SR for the processing of information in neural systems seems evident at all levels from the lower physiologieaf levels to the higher eognkil. e ones. Indeed, it has long been recognized that noise can improve the performance of certain neural networks [11],and it may be possible that an optimum noise level can achieve the maximum improvement.In this Letter, we consider a noisy autoassociative neural network which has previously been shown to be an accurate model of the bistable perceptual process involved in the interpretation of ambiguous figures [12]. Our results indicate that SR, as well as other recently studied f'eatures of noisy bistable dynamics, can easily be demonstrated in this system.Our network is made up of binary neurons (activation levels 0 and 1) which are globally, or all to all, connected [13]. The connection matrix is sy...
Two psychophysics experiments are described, pointing out the significant role played by stochastic resonance in recognition of capital stylized noisy letters by the human perceptive apparatus. The first experiment shows that an optimal noise level exists at which the letter is recognized for a minimum threshold contrast. A simple two-parameter model that best fits the experimental data is also discussed. In the second experiment we show that a dramatically increased ability of the visual system in letter recognition occurs in an extremely narrow range of increasing noise. Possible interesting future investigations suggested by these experimental results and based on functional imaging techniques are discussed.
Five probability distributions for the description of temporal fluctuations in the perception of ambiguous figures were fitted to previously obtained experimental results and classified according to their efficiency in describing the data. The gamma, Wiener, and Capocelli-Riciardi distributions showed the highest efficiency, while the chi2 and Taylor-Aldridge distributions showed a very low effiency. Therefore the underlying process may be described either by a simple Poisson model or by a random-walk model. For the gamma distribution there was a strong correlation between the parameters, while for the Wiener distribution this correlation was lower.
We verified whether a stochastic resonance paradigm (SR), with random interference (“noise”) added in optimal amounts, improves the detection of sub-threshold visual information by subjects with retinal disorder and impaired vision as it does in the normally sighted. Six levels of dynamic, zero-mean Gaussian noise were added to each pixel of images (13 contrast levels) in which alphabet characters were displayed against a uniform gray background. Images were presented with contrast below the subjective threshold to 14 visually impaired subjects (age: 22–53 yrs.). The fraction of recognized letters varied between 0 and 0.3 at baseline and increased in all subjects when noise was added in optimal amounts; peak recognition ranged between 0.2 and 0.8 at noise sigmas between 6 and 30 grey scale values (GSV) and decreased in all subjects at noise levels with sigma above 30 GSV. The results replicate in the visually impaired the facilitation of visual information processing with images presented in SR paradigms that has been documented in sighted subjects. The effect was obtained with low-level image manipulation and application appears readily possible: it would enhance the efficiency of today vision-improving aids and help in the development of the visual prostheses hopefully available in the future.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.