A composite model of the auditory periphery for simulating responses to complex sounds

Responses of midbrain auditory neurons to two different environmental sounds-A new approach on cross-sound modeling

When modeling auditory responses to environmental sounds, results are satisfactory if both training and testing are restricted to datasets of one type of sound. To predict 'cross-sound' responses (i.e., to predict the response to one type of sound e.g., rat Eating sound, after training with another type of sound e.g., rat Drinking sound), performance is typically poor. Here we implemented a novel approach to improve such cross-sound modeling (single unit datasets were collected at the auditory midbrain of anesthetized rats). The method had two key features: (a) population responses (e.g., average of 32 units) instead of responses of individual units were analyzed; and (b) the long sound segment was first divided into short segments (single sound-bouts), their similarity was then computed over a new metric involving the response (called Stimulus Response Model map or SRM map), and finally similar sound-bouts (regardless of sound type) and their associated responses (peri-stimulus time histograms, PSTHs) were modelled. Specifically, a committee machine model (artificial neural networks with 20 stratified spectral inputs) was trained with datasets from one sound type before predicting PSTH responses to another sound type. Model performance was markedly improved up to 92%. Results also suggested the involvement of different neural mechanisms in generating the early and late responses to amplitude transients in the broad-band environmental sounds. We concluded that it is possible to perform rather satisfactory cross-sound modeling on datasets grouped together based on their similarities in terms of the new metric of SRM map.

A hardware model of the auditory periphery to transduce acoustic signals into neural activity

To improve the performance of cochlear implants, we have integrated a microdevice into a model of the auditory periphery with the goal of creating a microprocessor. We constructed an artificial peripheral auditory system using a hybrid model in which polyvinylidene difluoride was used as a piezoelectric sensor to convert mechanical stimuli into electric signals. To produce frequency selectivity, the slit on a stainless steel base plate was designed such that the local resonance frequency of the membrane over the slit reflected the transfer function. In the acoustic sensor, electric signals were generated based on the piezoelectric effect from local stress in the membrane. The electrodes on the resonating plate produced relatively large electric output signals. The signals were fed into a computer model that mimicked some functions of inner hair cells, inner hair cell–auditory nerve synapses, and auditory nerve fibers. In general, the responses of the model to pure-tone burst and complex stimuli accurately represented the discharge rates of high-spontaneous-rate auditory nerve fibers across a range of frequencies greater than 1 kHz and middle to high sound pressure levels. Thus, the model provides a tool to understand information processing in the peripheral auditory system and a basic design for connecting artificial acoustic sensors to the peripheral auditory nervous system. Finally, we discuss the need for stimulus control with an appropriate model of the auditory periphery based on auditory brainstem responses that were electrically evoked by different temporal pulse patterns with the same pulse number.

A sound processor for cochlear implant using a simple dual path nonlinear model of basilar membrane

We propose a new active nonlinear model of the frequency response of the basilar membrane in biological cochlea called the simple dual path nonlinear (SDPN) model and a novel sound processing strategy for cochlear implants (CIs) based upon this model. The SDPN model was developed to utilize the advantages of the level-dependent frequency response characteristics of the basilar membrane for robust formant representation under noisy conditions. In comparison to the dual resonance nonlinear model (DRNL) which was previously proposed as an active nonlinear model of the basilar membrane, the SDPN model can reproduce similar level-dependent frequency responses with a much simpler structure and is thus better suited for incorporation into CI sound processors. By the analysis of dominant frequency component, it was confirmed that the formants of speech are more robustly represented after frequency decomposition by the nonlinear filterbank using SDPN, compared to a linear bandpass filter array which is used in conventional strategies. Acoustic simulation and hearing experiments in subjects with normal hearing showed that the proposed strategy results in better syllable recognition under speech-shaped noise compared to the conventional strategy based on fixed linear bandpass filters.

An improved speech processing strategy for cochlear implants based on an active nonlinear filterbank model of the biological cochlea

The purpose of this study was to improve the speech processing strategy for cochlear implants (CIs) based on a nonlinear time-varying filter model of a biological cochlea. The level-dependent frequency response characteristic of the basilar membrane is known to produce robust formant representation and speech perception in noise. A dual resonance nonlinear (DRNL) model was adopted because it is simpler than other adaptive nonlinear models of the basilar membrane and can be readily incorporated into the CI speech processor. Spectral analysis showed that formant information is more saliently represented at the output of the proposed CI speech processor compared to the conventional strategy in noisy conditions. Acoustic simulation and hearing experiments showed that the DRNL-based nonlinear strategy improves speech performance in a speech-spectrum-shaped noise.

Influence of inhibitory inputs on rate and timing of responses in the anteroventral cochlear nucleus

Anatomical and physiological studies have shown that anteroventral cochlear nucleus (AVCN) neurons receive glycinergic and GABAergic inhibitory inputs. In this study, changes in the temporal responses of AVCN neurons to pure tones and complex sounds after blocking inhibition were analyzed. Blocking inhibition influenced the temporal responses of each type of AVCN neuron. Choppers showed more chopping peaks and shortened chopping cycles after blocking inhibition. Sustained and slowly adapting choppers showed increased regularity throughout the response duration after blocking inhibition, whereas most transient choppers showed increased regularity in the early part of the response. Diverse changes in temporal response patterns were observed in neurons with primary-like and unusual responses, with several neurons showing a large decrease in the first-spike latency after blocking inhibition. This result disagreed with previous findings that onset responses are less affected than sustained responses by manipulating inhibition. Although blocking inhibition had a greater effect on spontaneous activity than that on tone-evoked activity, the change in spontaneous activity was less significant because of larger variability. In addition, for relatively high level masker noises, blocking inhibition had similar effects on responses to noise-alone and noise-plus-tone stimuli, in contrast with previous studies with low-level background noise. In general, inhibition had an enhancing effect on temporal contrast only for responses to amplitude-modulated tones, for which envelope synchrony was enhanced. Results of this study contribute new information about the characteristics, functional roles, and possible sources of inhibitory inputs received by AVCN neurons.

Further studies on the dual-resonance nonlinear filter model of cochlear frequency selectivity: responses to tones

A number of phenomenological models that simulate the response of the basilar membrane motion can reproduce a range of complex features observed in animal measurements over different sites along its cochlea. The present report shows a detailed analysis of the responses to tones of an improved model based on a dual-resonance nonlinear filter. The improvement consists in adding a third path formed by a linear gain and an all-pass filter. This improvement allows the model to reproduce the gain and phase plateaus observed empirically at frequencies above the best frequency. The middle ear was simulated by using a digital filter based on the empirical impulse response of the chinchilla stapes. The improved algorithm is evaluated against observations of basilar membrane responses to tones at seven different sites along the chinchilla cochlear partition. This is the first time that a whole set of animal observations using the same technique has been available in one species for modeling. The resulting model was able to simulate amplitude and phase responses to tones from basal to apical sites. Linear regression across the optimized parameters for seven different sites was used to generate a complete filterbank.

An improved speech processor for cochlear implant based on active nonlinear model of biological cochlea

The purpose of this study was to improve speech perception performance of cochlear implant (CI) under noise by a speech processing strategy based on nonlinear time-varying filter model of biological cochlea, which is beneficial in preserving spectral cues for speech perception. A dual resonance nonlinear model was applied to implement this feature. Time-frequency analysis indicated that formant information was more clearly represented at the output of CI speech processor, especially under noise. Acoustic simulation and hearing experiment also showed the superiority of the proposed strategy in that vowel perception score was notably enhanced. It was also observed that the AN responses to the stimulation pulses produced by the proposed strategy encode the formant information faithfully. Since the proposed strategy can be employed in CI devices without modification of hardwares, a significant contribution for the improvement of speech perception capability of CI implantees is expected.

Learning of auditory equivalence classes for vowels by rats

Four male Long-Evans rats were trained to discriminate between synthetic vowel sounds using a GO/NOGO response choice task. The vowels were characterized by an increase in fundamental frequency correlated with an upward shift in formant frequencies. In an initial phase we trained the subjects to discriminate between two vowel categories using two exemplars from each category. In a subsequent phase the ability of the rats to generalize the discrimination between the two categories was tested. To test whether rats might exploit the fact that attributes of training stimuli covaried, we used non-standard stimuli with a reversed relation between fundamental frequency and formants. The overall results demonstrate that rats are able to generalize the discrimination to new instances of the same vowels. We present evidence that the performance of the subjects depended on the relation between fundamental and formant frequencies that they had previously been exposed to. Simple simulation results with artificial neural networks could reproduce most of the behavioral results and support the hypothesis that equivalence classes for vowels are associated with an experience-driven process based on general properties of peripheral auditory coding mixed with elementary learning mechanisms. These results suggest that rats use spectral and temporal cues similarly to humans despite differences in basic auditory capabilities.

Modeling auditory-nerve responses for high sound pressure levels in the normal and impaired auditory periphery

This paper presents a computational model to simulate normal and impaired auditory-nerve (AN) fiber responses in cats. The model responses match physiological data over a wider dynamic range than previous auditory models. This is achieved by providing two modes of basilar membrane excitation to the inner hair cell (IHC) rather than one. The two modes are generated by two parallel filters, component 1 (C1) and component 2 (C2), and the outputs are subsequently transduced by two separate functions. The responses are then added and passed through the IHC low-pass filter followed by the IHC-AN synapse model and discharge generator. The C1 filter is a narrow-band, chirp filter with the gain and bandwidth controlled by a nonlinear feed-forward control path. This filter is responsible for low and moderate level responses. A linear, static, and broadly tuned C2 filter followed by a nonlinear, inverted and nonrectifying C2 transduction function is critical for producing transition region and high-level effects. Consistent with Kiang's two-factor cancellation hypothesis, the interaction between the two paths produces effects such as the C1/C2 transition and peak splitting in the period histogram. The model responses are consistent with a wide range of physiological data from both normal and impaired ears for stimuli presented at levels spanning the dynamic range of hearing.

Two New Directions in Speech Processor Design for Cochlear Implants

Two new approaches to the design of speech processors for cochlear implants are described. The first aims to represent "fine structure" or "fine frequency" information in a way that it can be perceived and used by patients, and the second aims to provide a closer mimicking than was previously possible of the signal processing that occurs in the normal cochlea.

Cochlear implants: some likely next steps

The history of cochlear implants is marked by large improvements in performance, especially over the past two decades and especially due to the development of ever-better processing strategies. Although the progress to date has been substantial, present devices still do not restore normal speech reception, even for top performers and particularly for listening to speech in competition with noise or other talkers. In addition, a wide range of outcomes persists, with some patients receiving little benefit using the same devices that support high levels of speech reception for others. The purpose of this review is to describe some likely possibilities for further improvement, including (a) combined electric and acoustic stimulation of the auditory system for patients with significant residual hearing, (b) use of bilateral implants, (c) a closer replication with implants of the processing steps in the normal cochlea, and (d) applications of knowledge about factors that are correlated with outcomes to help patients presently at the low end of the performance scale.

A phenomenological model for the responses of auditory-nerve fibers. II. Nonlinear tuning with a frequency glide

A computational model was developed to simulate the responses of auditory-nerve (AN) fibers in cat. The model's signal path consisted of a time-varying bandpass filter; the bandwidth and gain of the signal path were controlled by a nonlinear feed-forward control path. This model produced realistic response features to several stimuli, including pure tones, two-tone combinations, wideband noise, and clicks. Instantaneous frequency glides in the reverse-correlation (revcor) function of the model's response to broadband noise were achieved by carefully restricting the locations of the poles and zeros of the bandpass filter. The pole locations were continuously varied as a function of time by the control signal to change the gain and bandwidth of the signal path, but the instantaneous frequency profile in the revcor function was independent of sound pressure level, consistent with physiological data. In addition, this model has other important properties, such as nonlinear compression, two-tone suppression, and reasonable Q10 values for tuning curves. The incorporation of both the level-independent frequency glide and the level-dependent compressive nonlinearity into a phenomenological model for the AN was the primary focus of this work. The ability of this model to process arbitrary sound inputs makes it a useful tool for studying peripheral auditory processing.

A non-linear circuit for simulating OHC of the cochlea

In the present paper, referring to known characteristics of the outer hair cells functioning in the cochlea of the inner ear, a functional model of the outer hair cells is constructed. It consists of a linear feed-forward circuit and a non-linear positive feedback circuit. The feed-forward circuit reflects the contribution of local basilar and tectorial membrane areas and passive outer hair cells' physical parameters to the forming of low-selectivity resonance characteristics. The non-linear positive feedback circuit reflects the non-linear outer hair cell signal transduction processes and the active role of efferents from the medial superior olive in altering circuit sensitivity and selectivity. Referring to an analytical description of the circuit model and computer simulation results, an explanation is given over the biological meaning of the outer hair cells' non-linearities in signal transduction processes and the role of the non-linearities in achieving the following: signal compression, the dependency of circuit sensitivity and frequency selectivity upon the input signal amplitude, the compatibility of high-frequency selectivity and short transient response of the biological filtering circuits.

A nonlinear filter-bank model of the guinea-pig cochlear nerve: rate responses

The aim of this study is to produce a functional model of the auditory nerve (AN) response of the guinea-pig that reproduces a wide range of important responses to auditory stimulation. The model is intended for use as an input to larger scale models of auditory processing in the brain-stem. A dual-resonance nonlinear filter architecture is used to reproduce the mechanical tuning of the cochlea. Transduction to the activity on the AN is accomplished with a recently proposed model of the inner-hair-cell. Together, these models have been shown to be able to reproduce the response of high-, medium-, and low-spontaneous rate fibers from the guinea-pig AN at high best frequencies (BFs). In this study we generate parameters that allow us to fit the AN model to data from a wide range of BFs. By varying the characteristics of the mechanical filtering as a function of the BF it was possible to reproduce the BF dependence of frequency-threshold tuning curves, AN rate-intensity functions at and away from BF, compression of the basilar membrane at BF as inferred from AN responses, and AN iso-intensity functions. The model is a convenient computational tool for the simulation of the range of nonlinear tuning and rate-responses found across the length of the guinea-pig cochlear nerve.

An auditory-periphery model of the effects of acoustic trauma on auditory nerve responses

Acoustic trauma degrades the auditory nerve's tonotopic representation of acoustic stimuli. Recent physiological studies have quantified the degradation in responses to the vowel /E/ and have investigated amplification schemes designed to restore a more correct tonotopic representation than is achieved with conventional hearing aids. However, it is difficult from the data to quantify how much different aspects of the cochlear pathology contribute to the impaired responses. Furthermore, extensive experimental testing of potential hearing aids is infeasible. Here, both of these concerns are addressed by developing models of the normal and impaired auditory peripheries that are tested against a wide range of physiological data. The effects of both outer and inner hair cell status on model predictions of the vowel data were investigated. The modeling results indicate that impairment of both outer and inner hair cells contribute to degradation in the tonotopic representation of the formant frequencies in the auditory nerve. Additionally, the model is able to predict the effects of frequency-shaping amplification on auditory nerve responses, indicating the model's potential suitability for more rapid development and testing of hearing aid schemes.

A revised model of the inner-hair cell and auditory-nerve complex

A revised computational model of the inner-hair cell (IHC) and auditory-nerve (AN) complex is presented and evaluated. Building on previous models, the algorithm is intended as a component for use in more comprehensive models of the auditory periphery. It combines smaller components that aim to be faithful to physiology in so far as is practicable and known. Transduction between cochlear mechanical motion and IHC receptor potential (RP) is simulated using a modification of an existing biophysical IHC model. Changes in RP control the opening of calcium ion channels near the synapse, and local calcium levels determine the probability of the release of neurotransmitter. AN adaptation results from transmitter depletion. The exact timing of AN action potentials is determined by the quantal and stochastic release of neurotransmitter into the cleft. The model reproduces a wide range of animal RP and AN observations. When the input to the model is taken from a suitably nonlinear simulation of the motion of the cochlear partition, the new algorithm is able to simulate the rate-intensity functions of low-, medium-, and high-spontaneous rate AN fibers in response to stimulation both at best frequency and at other frequencies. The variation in fiber type arises in large part from the manipulation of a single parameter in the model: maximum calcium conductance. The model also reproduces quantitatively phase-locking characteristics, relative refractory effects, mean-to-variance ratio, and first- and second-order discharge history effects.

Evaluating auditory performance limits: i. one-parameter discrimination using a computational model for the auditory nerve

A method for calculating psychophysical performance limits based on stochastic neural responses is introduced and compared to previous analytical methods for evaluating auditory discrimination of tone frequency and level. The method uses signal detection theory and a computational model for a population of auditory nerve (AN) fiber responses. The use of computational models allows predictions to be made over a wider parameter range and with more complete descriptions of AN responses than in analytical models. Performance based on AN discharge times (all-information) is compared to performance based only on discharge counts (rate-place). After the method is verified over the range of parameters for which previous analytical models are applicable, the parameter space is then extended. For example, a computational model of AN activity that extends to high frequencies is used to explore the common belief that rate-place information is responsible for frequency encoding at high frequencies due to the rolloff in AN phase locking above 2 kHz. This rolloff is thought to eliminate temporal information at high frequencies. Contrary to this belief, results of this analysis show that rate-place predictions for frequency discrimination are inconsistent with human performance in the dependence on frequency for high frequencies and that there is significant temporal information in the AN up to at least 10 kHz. In fact, the all-information predictions match the functional dependence of human performance on frequency, although optimal performance is much better than human performance. The use of computational AN models in this study provides new constraints on hypotheses of neural encoding of frequency in the auditory system; however, the method is limited to simple tasks with deterministic stimuli. A companion article in this issue ("Evaluating Auditory Performance Limits: II") describes an extension of this approach to more complex tasks that include random variation of one parameter, for example, random-level variation, which is often used in psychophysics to test neural encoding hypotheses.

A computational algorithm for computing nonlinear auditory frequency selectivity

Computational algorithms that mimic the response of the basilar membrane must be capable of reproducing a range of complex features that are characteristic of the animal observations. These include complex input output functions that are nonlinear near the site's best frequency, but linear elsewhere. This nonlinearity is critical when using the output of the algorithm as the input to models of inner hair cell function and subsequent auditory-nerve models of low- and high-spontaneous rate fibers. We present an algorithm that uses two processing units operating in parallel: one linear and the other compressively nonlinear. The output from the algorithm is the sum of the outputs of the linear and nonlinear processing units. Input to the algorithm is stapes motion and output represents basilar membrane motion. The algorithm is evaluated against published chinchilla and guinea pig observations of basilar membrane and Reissner's membrane motion made using laser velocimetry. The algorithm simulates both quantitatively and qualitatively, differences in input/output functions among three different sites along the cochlear partition. It also simulates quantitatively and qualitatively a range of phenomena including isovelocity functions, phase response, two-tone suppression, impulse response, and distortion products. The algorithm is potentially suitable for development as a bank of filters, for use in more comprehensive models of the peripheral auditory system.

A computational model of mechanisms controlling experience-dependent reorganization of representational maps in auditory cortex

Cortical representations of sound can be modified by repeatedly pairing presentation of a pure tone with electrical stimulation of neuromodulatory neurons located in the basal forebrain (Bakin & Weinberger, 1996; Kilgard & Merzenich, 1998a). We developed a computational model to investigate the possible effects of basal forebrain modulation on map reorganization in the auditory cortex. The model is a self-organizing map with acoustic response characteristics mimicking those observed in the mammalian auditory cortex. We simulated the effects of basal forebrain modulation, using parameters intrinsic to the self-organizing map, such as the learning rate (controlling the adaptability of map nodes) and the neighborhood function (controlling the excitability of map nodes). Previous research has suggested that both parameters can be useful for characterizing the effects of neuromodulation on plasticity (Kohonen, 1993; Myers et al., 1996; Myers, Ermita, Hasselmo, & Gluck, 1998). The model successfully accounts for experimentally observed effects of pairing basal forebrain stimulation with the presentation of a single tone, but not of two tones, suggesting that auditory cortical plasticity is constrained in ways not accounted for by current theories. Despite this limitation, the model provides a useful framework for describing experience-induced changes in auditory representations and for relating such changes to variations in the excitability and adaptability of cortical neurons produced by neuromodulation.

A phenomenological model for the responses of auditory-nerve fibers: I. Nonlinear tuning with compression and suppression

A phenomenological model was developed to describe responses of high-spontaneous-rate auditory-nerve (AN) fibers, including several nonlinear response properties. Level-dependent gain (compression), bandwidth, and phase properties were implemented with a control path that varied the gain and bandwidth of tuning in the signal-path filter. By making the bandwidth of the control path broad with respect to the signal path, the wide frequency range of two-tone suppression was included. By making the control-path filter level dependent and tuned to a frequency slightly higher than the signal-path filter, other properties of two-tone suppression were also included. These properties included the asymmetrical growth of suppression above and below the characteristic frequency and the frequency offset of the suppression tuning curve with respect to the excitatory tuning curve. The implementation of this model represents a relatively simple phenomenological description of a single mechanism that underlies several important nonlinear response properties of AN fibers. The model provides a tool for studying the roles of these nonlinearities in the encoding of simple and complex sounds in the responses of populations of AN fibers.

The representation of pure tones and noise in a model of cochlear nucleus neurons

The limited dynamic range of the majority of auditory-nerve fibers represents a difficulty in accounting for normal hearing capabilities over the known psychoacoustic intensity range. The presence of noise is an additional complication because it will tend to saturate these fibers, thereby considerably reducing their dynamic range, i.e., the range of mean firing rates. In this study, simulations involving a model of auditory nerve and cochlear nucleus neurons were conducted using pure-tone stimuli in the presence of noise. The main focus is on the role of inhibition in regulating the activity of cells, improving their capability to represent signals in background noise. This concerns in particular those inhibitory neurons that receive input from a wide range of auditory-nerve fibers and respond with an onset chopper pattern. A detailed model of stellate cells is used. It allows several parameters such as the number, location, and strength of inputs to be manipulated. The fist part of this paper presents the model and its responses to pure-tone and noise stimuli presented separately. The model's capacity to generalize to tone/noise combinations is then tested. Responses to these stimuli are found to be qualitatively similar to neurophysiological findings. Model neurons exhibit appropriate shifts in their rate-level functions and their responses are inhibited or suppressed by tones outside their characteristic frequency. The model stellate cell is also found to display many of the temporal patterns reported in electrophysiological studies as a result of appropriate settings of certain parameters. Therefore, the model is sufficient to account for a larger number of findings and should serve as a basis for predicting responses to novel stimuli, or as a building block for modeling larger networks.