Browsing by Author "Shamma, S.A."
Results Per Page
Sort Options
Item Analysis of Dynamic Spectra in Ferret Primary Auditory Cortex: I. Characteristics of Single Unit Responses to Moving Ripple Spectra(1995) Kowalski, Nina; Depireux, Didier A.; Shamma, S.A.; ISRAuditory stimuli referred to as moving ripples are used to characterize the responses of both single and multiple units in the ferret primary auditory cortex (AI). Moving ripples are broadband complex sounds with sinusoidal spectral profiles that drift along the tonotopic axis at a constant velocity. Neuronal responses to moving ripples are locked to the phase of the ripple, i.e., they exhibit the same periodicity as that of the moving ripple profile. Neural responses are characterized as a function of ripple velocity (temporal property) and ripple frequency (spectral property). Transfer functions describing the response to these temporal and spectral modulations are constructed. Temporal transfer functions are inverse Fourier transformed to obtain impulse response functions that reflect the cell's temporal characteristics. Ripple transfer functions are inverse Fourier transformed to obtain the response field, characterizing the cell's response area along the tonotopic axis. These operations assume linearity in the cell's response to moving ripples. Separability of the temporal and ripple transfer functions is established by comparing transfer functions across different test parameters. Response fields measured with either stationary ripples or moving ripples are shown to be similar. Separability implies that the neuron can be modeled as processing spatio-temporal information in two distinct stages. The assumption of linearity implies that each of these stages is a linear operation.The ripples parameters that characterize cortical cells are distributed somewhat evenly, with the characteristic ripple frequencies ranging from 0.2 to over 2 cycles/octave and the characteristic angular frequency typically ranging from 2 to 20 Hz. Many responses exhibit periodicities not found in the spectral envelope of the stimulus. These periodicities are of two types. Slow rebounds with a period of about 150 ms appear with various strengths in about 30 % of the cells. Fast regular firings, with interspike intervals of the order of 10 ms are much less common and may reflect the ability of certain cells to follow the fine structure of the stimulus.
Item Analysis of Dynamic Spectra in Ferret Primary Auditory Cortex: II. Prediction of, Unit Responses to Arbitrary Dynamic Spectra(1995) Kowalski, Nina; Depireux, Didier A.; Shamma, S.A.; ISRResponses of single units and unit clusters were recorded in the ferret primary auditory cortex (AI) using broadband complex dynamic spectra. Previous work (Kowalski et al 1995) demonstrated that simpler spectra consisting of single moving ripples (i.e., sinusoidally modulated spectral profiles that travel at a constant velocity along the logarithmic frequency axis) could be used effectively to characterize the response fields and transfer functions of AI cells. An arbitrary complex dynamic spectral profile can be thought of conceptually as being composed of a weighted sum of moving ripple spectra. Such a decomposition can be computed from a two-dimensional spectro- temporal Fourier transform of the dynamic spectral profile with moving ripples as the basis function. Therefore, if AI units were essentially linear satisfying the superposition principle, then their responses to arbitrary dynamic spectra could be predicted from the responses to single moving ripples, i.e., from the units response fields and transfer functions. This conjecture was tested and confirmed with data from 293 combinations of moving ripples, involving complex spectra composed of up to 15 moving ripples of different ripple frequencies and velocities. For each case, response predictions based on the unit transfer functions were compared to measured responses. The correlation between predicted and measured responses was found to be consistently high (84% with rho > 0.6). The distribution of response parameters suggest that AI cells may encode the profile of a dynamic spectrum by performing a multiscale spectro-temporal decomposition of the dynamic spectral profile in a largely linear manner.Item The Auditory Processing of Speech.(1986) Shamma, S.A.; ISRThe processing of speech in the mammalian auditory periphery is discussed in terms of the spatio-temporal nature of the distribution of the cochlear response and the novel encoding schemes this permits. Algorithms to detect specific morphological features of the response and the novel encoding schemes of the response patterns are also considered for the extraction of stimulus spectral parameters.Item Auditory Representations of Acoustic Signals(1991) Yang, X.; Wang, K.; Shamma, S.A.; ISRAn analytically tractable framework is presented to describe neural processing in the early stages of the auditory system. Algorithms are developed to assess the integrity of the acoustic spectrum at all processing stages. The algorithms employ wavelet representations, multiresolution processing, and the method of convex projections to reconstruct close replica of the input stimulus. Reconstructions using natural speech sounds demonstrate minimal loss of information along the auditory pathway. Furthermore, close inspections of the final auditory patterns reveals spectral enhancements and noise suppression that have close perceptual correlates. Finally, the auditory representations are shown to be versatile for many applications, including automatic speech recognition and low bit-rate data compression.Item Dynamics of Neural Responses in Ferret Primary Auditory Cortex: I. Spectro-Temporal Response Field Characterization by Dynamic Ripple Spectra(1999) Depireux, Didier A.; Simon, J.Z.; Klein, David J.; Shamma, S.A.; ISR; CAARTo understand the neural representation of broadband, dynamic sounds in Primary Auditory Cortex (AI), we characterize responses using the Spectro-Temporal Response Field (STRF). The STRF describes and predicts the linear response of neurons to sounds with rich spectro-temporal envelopes. It is calculated here from the responses to elementary "ripples," a family of sounds with drifting, sinusoidal, spectral envelopes--the complex spectro-temporal envelope of any broadband, dynamic sound can expressed as the linear sum of individual ripples.The collection of responses to all elementary ripples is the spectro-temporal transfer function. Previous experiments using ripples with downward drifting spectra suggested that the transfer function is separable, i.e., it is reducible into a product of purely temporal and purely spectral functions.
Here we compare the responses to upward and downward drifting ripples, assuming separability within each direction, to determine if the total bi-directional transfer function is fully separable. In general, the combined transfer function for two directions is not symmetric, and hence units in AI are not, in general, fully separable. Consequently, many AI units have complex response properties such as sensitivity to direction of motion, though most inseparable units are not strongly directionally selective.
We show that for most neurons the lack of full separability stems from differences between the upward and downward spectral cross-sections, not from the temporal cross-sections; this places strong constraints on the neural inputs of these AI units.
Item Identification of Connectiviity in Neural Networks.(1989) Yang, X.; Shamma, S.A.; ISRAnalytical and experimental methods are provided for estimating synaptic connectivities from simultaneous recordings of multiple neurons. The results are based on detailed, yet flexible neuron models in which spike trains are modeled as general doubly stochastic point processes. The expressions derived can be used with non-stationary or stationary records, and can be readily extended from pair-wise to multi-neuron estimates. Furthermore, we show analytically how the estimates are improved as more neurons are sampled, and derive the appropriate normalizations to eliminate stimulus-related correlations. Finally, we illustrate the use and interpretation of the analytical expressions on simulated spike trains and neural networks, and give explicit confidence measures on the estimates.Item Minimum Mean Square Error Estimation of Connectivity in Biological Neural Networks(1991) Yang, X.; Shamma, S.A.; ISRA minimum mean square error (MMSE) estimation scheme is employed to identify the synaptic connectivity in neural networks. This new approach can substantially reduce the amount of data and the computational cost involved in the conventional correlation methods, and is suitable for both nonstationary and stationary neuronal firings. Two algorithms are proposed to estimate the synaptic connectivities recursively, one for nonlinear filtering, the other for linear filtering. In addition, the lower and upper bounds for the MMSE estimator are determined. It is shown that the estimators are consistent in quadratic mean. We also demonstrate that the conventional crossinterval histogram is an asymptotic linear MMSE estimator with an inappropriate initial value. Finally, simulations of both the nonlinear and linear (Kalman filter) estimates demonstrate that the true connectivity values are approached asymptotically.Item Neural Networks for Speech Processing and Recognition.(1987) Shamma, S.A.; ISRIn this report, we shall outline a specific approach to the analysis and recognition of speech phonemes based on the fundamental principles of processing in the auditory nervous system. The system consists of particular implementations of the three conceptual stages mentioned above: The cochlear transformations of speech sounds into spatiotemporal patterns (stage 1); the subsequent feature extraction by the central neural networks (stage 2); and the use of various adaptive nets to identify the acoustic features of speech phonemes (stage 3). We shall illustrate the utility of this approach in identifying and organizing the invariant features of nine American English vowels.Item On the Prediction of Local Patterns in Cellular Automata.(1986) Wilbur, W.J.; Lipman, David J.; Shamma, S.A.; ISRThe class of deterministic one-dimensional cellular automata studied recently by Wolfram are considered. We represent a state of an automaton as a probability distribution of patterns of a fixed size. In this way information is lost but it is possible to approximate the stepwise action of the automaton by the iteration of an analytic mapping of the set of probability distributions to itself. Such nonlinear analytic mappings generally have nontrivial attractors and in the most interesting cases (Wolfram Class III) these are single points. The point attractors under appropriate circumstances provide good approximations to the frequencies of local patterns generated by the discrete rules from which they were derived. Two appropriate settings for such approximations are transient patterns generated from random starts and patterns generated in a noisy environment. In the case with noise, improvement is found by correction of the analytic mappings for the effects of noise. Examples of both types of approximations are considered.Item Representation of Spectral Profiles in the Auditory System Part II: A Ripple Analysis Model(1993) Vranic-Sowers, S.; Shamma, S.A.; ISRBased on experimental results presented in [Vranic-Sowers and Shamma, 1993], and on further physiological and psychoacoustical evidence, it is argued that the auditory system analyzes a spectral profile along two largely independent dimensions. They correspond to the magnitude and phase of a localized Fourier transformation of the profile, closely analogues to the spatial frequency transformations described in the visual system. Within this general framework, a model of profile analysis is proposed in which a spectral profile is assumed to be represented by a weighted sum of sinusoidally modulated spectra (ripples). The analysis is performed by a bank of bandpass filters, each tuned to a particular ripple frequency and ripple phase. The parameters of the model are estimated using data from ripple detection experiments in [Green, 1986; Hillier, 1991]. Perceptual thresholds are then computed from the filter outputs and compared with thresholds measured for peak profile experiments, and for detection tasks with step, single component increment, and the alternating profiles.Item Representation of Spectral Profiles in the Auditory System, I: A Ripple Analysis Model(1994) Vranic-Sowers, S.; Shamma, S.A.; ISRA model of profile analysis is proposed in which a spectral profile is assumed to be represented by a weighted sum of sinusoidally modulated spectra (ripples). The analysis is performed by a bank of bandpass filters, each tuned to a particular ripple frequency and ripple phase. The parameters of the model are estimated using data from ripple detection experiments in [Green} 1986; Hillier 1991]. Detection thresholds are computed from the filter outputs and compared with perceptual thresholds, for profile detection experiments with step, single component increment, and the alternating profiles. The model accounts well for the measured thresholds in these experiments. Physiological and psychophysical evidences from the auditory and visual systems in support of this type of a model are also reviewed. The implications of this model for pitch and timbre perception are briefly discussed.Item Representation of Spectral Profiles in the Auditory System, II: Detection of Spectral Peak Shape Changes(1994) Vranic-Sowers, S.; Shamma, S.A.; ISRBased on the ripple analysis model outlined in Part I of this paper, predictions are made for the detection of shape changes in spectral peak profiles. Peak shape is uniquely described in terms of two parameters: bandwidth factor (BWF) which reflects the tuning or sharpness of a peak, and a symmetry factor (SF) which roughly measures the local evenness or oddness of a peak. Using profile analysis methods, thresholds to changes in these parameters (defined as dBWF/BWF and dSF) are measured together with the effects of several manipulations such as using different peak levels, varying spectral component densities, and randomizing the frequencies of the peaks. The new ripple analysis model accounts well for the measured thresholds. Predictions of the three previously published models for the same profiles are also evaluated and discussed.Item Representation of Spectral Profiles in the Auditory Systems(1992) Vranic, S.; Shamma, S.A.; ISRThis paper explores the question of how spectral profiles (such as spectral peaks) might be represented and perceived in the auditory system. Using profile analysis methods, we measured listeners' sensitivities to changes in spectral peak shapes that were uniquely described in terms of two parameters: a symmetry factor (SF) which roughly measure the local evenness or oddness of a peak, and a bandwidth factor (BWF) which reflects the tuning or sharpness of a peak. The effects of several manipulations on the perceptual thresholds were also tested; they include using different peak levels, varying spectral component densities, and randomizing the frequencies of the peaks. The basic result that emerges is that threasholds to changes in SF and BWF are constant regardless of peak shape. Thus, for the detection of SF changes, dSF thresholds are approximately constant regardless of a peak's SF and BWF. The only exception occurs towards the narrowest peaks where detection thresholds rise. For the detection of BWF changes, all dBWF/BWF thresholds remain constant regardless of peak shape. A fundamental conclusion arising from these data is that peak profiles are represented along two sensitive and largely independent axes: peak bandwidth and symmetry factors. More generally, however, it is argued that for an arbitrary spectral profile these two axes simply correspond to the magnitude and phase of a fourier transformation (or more precisely, of a Wavelet transformation) of the profile, closely analogous to the spatial frequency transformations described in the visual system. Further physiological and phychophysical evidence in support of this hypothesis is discussed.Item Ripple Analysis in Ferret Primary Auditory Cortex. II. Topographic and Columnar Distribution of Ripple Response Parameters(1994) Versnel, H.; Kowalski, Nina; Shamma, S.A.; ISRWe examined the columnar and topographic distribution of response parameters using spectral ripples and tonal stimuli in the primary auditory cortex (AI) of the barbiturate-anesthetized ferret. The ripple stimuli consisted of broadband stimuli (1-20 kHz) with sinusoidally modulated spectral envelopes.Responses to ripples were parametrized in terms of characteristic ripple Wo(ripple frequency where the magnitude of the ripple transfer function is maximal, i.e., where the cell responds best) and characteristic phase Fo (intercept of the phase of the ripple transfer function, i.e., phase where the cell responds best). The response area (measured with tones) was parametrized in terms of its excitatory bandwidth at 20 dB above threshold (BW20), and its asymmetry as reflected by the directional sensitivity index (C) to frequency-modulated (FM) tones. Columnar organization for the above four parameters was investigated in 66 single units from 23 penetrations. It was confirmed for Wo, Fo, and the C index, but it appeared to be ambiguous for BW20. The response parameters measured from multiunit recordings corresponded closely to those obtained from single units in the same cluster. In a local region, most cells exhibited closely matched, response fields (RFs, inverse Fourier transformed ripple transfer function) and response areas (measured with two-tone stimuli), and had correspondingly similar response parameters to ripples and tones. The topographic distribution of the response parameters across the surface of AI was studied with multiunit recordings in four animals. In all maps, systematic patterns or clustering of, response parameters could be discerned along the isofrequency planes.
The distribution of the characteristic ripple Wo exhibited two trends. First, along the isofrequency planes, it was largest near the center of AI, gradually decreasing towards the edges of the field where often a secondary maximum was found.
The second trend occurred along the tonotopic axis where the maximum Wo found in an isofrequency range increases with increasing BF. The tonal bandwidth BW20, which was inversely correlated with Wo, exhibited a similar topographic distribution along the tonotopic axis and the isofrequency planes. The distribution of the characteristic ripple phase, Fo which reflects the asymmetry in the response field, showed a systematic order along the isofrequency axis. At the center of AI symmetric responses (Fo 0) predominated. Towards the edges, the RFs became more asymmetric with Fo < 0 caudally, and Fo > 0 rostrally. The asymmetric response types tended to cluster along repeated bands that paralleled the tonotopic axis. The FM directional sensitivity (C index, reflecting asymmetry of tonal response areas) tends to have similar trends along the isofrequency axis as Fo.
Item Robust Spectro-Temporal Reverse Correlation for the Auditory System: Optimizing Stimulus Design(1999) Klein, David J.; Depireux, Didier A.; Simon, J.Z.; Shamma, S.A.; ISR; CAARThe spectro-temporal receptive field (STRF) is a functionaldescriptor of the linear processing of time-varying acoustic spectra by theauditory system. By cross-correlating sustained neuronal activity with the"dynamic spectrum" of a spectro-temporally rich stimulus ensemble, oneobtains an estimate of the STRF.In this paper, the relationship betweenthe spectro-temporal structure of any given stimulus and the quality ofthe STRF estimate is explored and exploited. Invoking the Fouriertheorem, arbitrary dynamic spectra are described as sums of basicsinusoidal components, i.e., "moving ripples." Accurate estimation isfound to be especially reliant on the prominence of components whosespectral and temporal characteristics are of relevance to the auditorylocus under study, and is sensitive to the phase relationships betweencomponents with identical temporal signatures.
These and otherobservations have guided the development and use of stimuli withdeterministic dynamic spectra composed of the superposition of many"temporally orthogonal" moving ripples having a restricted, relevant rangeof spectral scales and temporal rates.
The method, termedsum-of-ripples, is similar in spirit to the "white-noise approach," butenjoys the same practical advantages--which equate to faster and moreaccurate estimation--attributable to the time-domain sum-of-sinusoidsmethod previously employed in vision research. Application of the methodis exemplified with both modeled data and experimental data from ferretprimary auditory cortex (AI).
Item The Second Synthetic Microstructures in Biological Research Conference.(1988) Peckerar, M.; Krishnaprasad, Perinkulam S.; Shamma, S.A.; ISRNot available.Item Spectral Gradient Columns in Primary Auditory Cortex: Physiological and Psychoacoustical Correlates(1991) Shamma, S.A.; Vranic, S.; Wiser, P.; ISRMapping of the spatial distribution of responses in the primary auditory cortex (AI) reveal that both the gradient of the acoustic spectrum and sensitivity to FM sweep direction are encoded in an orderly manner along the isofrequency planes of AI. Psychoacoustical tests also demonstrate a potential perceptual correlate of the gradient maps, namely the threshold stability and heightened sensitivity of human subjects to the detection of changes in the symmetry of spectral peaks.Item Synchrony Suppression in Complex Stimulus Responses of Biophysical Model of the Cochlea.(1986) Shamma, S.A.; Morrish, K.A.; ISRA minimal biophysical model of the cochlea is used to investigate the validity of the hypothesis that a single compressive nonlinearity at the hair cell level can explain some of the synchrony suppression phenomena in cochlear response to complex stimuli. The dependencies of the model responses on the amplitudes and frequencies of two-tone stimuli resemble in many respects the behavior of the experimental data, and can be traced to explicit biophysical parameters in the model. Most discrepancies between theory and experiment stem from simplifications in parameters of the minimal model that play no direct role in the hypothesis. The analysis and simulations predict further results which, pending experimental verification, may provide a more direct test of the influence of the compressive nonlinearity on the relative amplitudes of the synchronous response components, and hence of its role in synchrony suppression. For instance, regardless of the overall absolute levels of a two-tone stimulus applied to this type of model, the ratio of the amplitudes at the input, and the ratio of the corresponding responses at the output, remain approximately constant and equal (the output ratio changes by at most 6 dB in favor of the stronger tone). Other nonlinear responses to multi- tonal stimuli can also be reproduced such as 'spectral edge enhancement' (Horst et al. [l985]), Springer-Verlag) and some aspects of three-tone suppression (Javel et al. [l983]), Monash University Press). In contrast to the complex behavior of the synchrony suppression with increasing intensity, and the resulting drastic influences of the compressive nonlinear on the response measures on the auditory-nerve (e.g. average rate and synchrony profiles), the percepts of complex sounds are relatively stable. This suggests that the invariant response measures are more likely used in the encoding and CNS extraction of the spectrum of complex stimuli such as speech.Item A Totally Automated Neural Spike Detection and Classification Scheme: A Preliminary Software System.(1986) Shamma, S.A.; Yang, X.; ISRA system for neural spike detection and classification is presented, which does not require a priori assumptions about spike presence or spike templates, and assumes only that the background noise has a Gaussian distribution. The system is divided into two parts: a learning subsystem and a real-time detection and classification subsystem. The former extracts templates of spikes for every class which includes a feature learning phase and a template learning phase. The latter picks up spikes in the noisy trace and sorts them out into classes, based on the templates that the learning subsystem provides and the statistics of the background noise. Performance of the system is illustrated by using it to classify spikes in a segment of neural activity recorded from monkey motor cortex. The system is implemented without human supervision so that it can be extended for multi-channel recording without loss of real-time property.Item Zero-Crossing and Noise Suppression in Auditory Wavelet Transformations(1992) Wang, K.; Shamma, S.A.; ISRA common sequence of operations in the early stages of most biological sensory systems is a wavelet transform followed by a compressive nonlinearity. In this paper, we explore the contribution of these operations to the formation of robust and perceptually significant representations in the auditory system. It is demonstrated that the neural representation of a complex signal such as speech is derived from a highly reduced version of its wavelet transform, specifically, from the distribution of its locally averaged zero-crossing rates along the temporal and scale axes. It is shown analytically that such encoding of the wavelet transform results in mutual suppressive interactions across its different scale representations. Suppression in turn endows the representation with enhanced spectral peaks and superior robustness in noisy environments. Examples using natural speech vowels are presented to illustrate the results. Finally, we discuss the relevance of these findings to conventional subband coding of speech signals.