Elsevier

Neurocomputing

Volume 72, Issues 4–6, January 2009, Pages 936-943
Neurocomputing

Low rate stochastic strategy for cochlear implants

https://doi.org/10.1016/j.neucom.2008.06.021Get rights and content

Abstract

In this article, a variant on low rate stochastic stimulation is presented. This variant shows more physiological characteristics, which are able to help preserving the cochlea due to low rate and low level stimulation. The speech processor here presented is provided with characteristics more similar to the normal peripheral auditory system. From the processor's output, neural stimuli, it is possible to estimate both sound intensity per bands or areas of frequency, and the value of the frequencies composing the sound. When there is no sound stimulus, the peripheral auditory system generates stochastic stimuli from which it is not possible to extract information on amplitude neither frequency. In this case of clearly random generation of stimuli, the spectrum is noisy on the whole frequency.

Introduction

Cochlear implants are devices used as a means to restore hearing in deep or total deaf patients. A cochlear implant stimulates directly the cochlea with electrical stimuli capable of evoking auditory sensation on a perceptive level.

Cochlear implants are classified according to the strategy used for speech processing (simultaneous analog stimulation, continuous interleaved sampler, F0/F2 and F0/F1/F2, ACE, PACE, etc) [3], [5], [22], [23], [24], [25], [26], [27], [28], the type of stimulation (analog or pulses), the signal distribution in time (simultaneous or sequential) and the position of the reference electrode (monopolar or bipolar).

As cochlear implants do not perform an optimum encoding of sound, it is necessary to research on new ways or variants of stimulation closer to the patterns of electric activity in the auditory nerve, so that the implant's behavior be as similar as possible to the natural or physiological normal ear.

Cochlear implants operating at high rate and high level stimulation produce high discharge probabilities in auditory nerve fiber (ANF) [7]. This may lead to an improved detection of the fine temporal components in speech among cochlear implant patients, but such stimulation could evoke metabolic stress on auditory nerve and may lead to neural degeneration [10].

The aim of this work is to prove that a speech processor based on the computational model of the auditory peripheral system [4] is able to provide more information about the perceived sound to the central nervous system (CNS), so it will bring a better performance to the tasks of speech and speaker recognition. The stimulation strategy is based on the spatiotemporal coding [11]; a set of electrodes stimulate a specific zone, which is related to a resonance frequency of the basilar membrane.

This variant provides a low rate stochastic stimulation allowing the possibility of stimulating with the minimum current levels for evoking auditory sensation and the advantage of a better protection of the cochlea. This variant also presents more physiological characteristics, because the information is not transported in the intensity of the applied electric stimulus but in the stimulation rate.

In the well-known artificial stimulation methods, the neuron possesses the control in the decision of the neural discharge, which shows a spike probability directly proportional to the electric current intensity of the stimulus applied to the neuron [7], [8], [9]. On the contrary, in the current stimulation proposal, this neuron's function is taken over by the ANF block of the speech processor which will decide the exact moment that the neuron is discharged. For this purpose, the electric stimulus intensity will have to be such that it provokes the neuron discharge in around 100% of the cases. On the other hand, the electric stimuli will always have to be of the same magnitude, and the neuron becomes a simple node on the way of the neuronal stimulus.

Section snippets

Brief description of the peripheral auditory system operation

The function of the ear in the human auditory system is the transduction of the sound waves in a pattern of electric activity in the auditory nerve, whose electric activity is processed by the brain in order to produce the perception of sound [3], [12].

The auditory sensorial loss refers to the interruption in the transduction of the sound waves in the neuronal activity of the auditory nerve, as a result of malformations or degeneration of Corti's organ within the cochlea.

In the organ of Corti,

Implementation of the speech processor

The speech processor (see Fig. 1) is based on the computational model of the auditory peripheral system presented in [4], which shows a nonlinear processing and a low rate stochastic stimulation of the stimulation of the cochlea, like a normal ear.

The function of the external and middle ear is to capture the sound waves and perform an initial filtering of the signal in order to increase the pressure of sound in the region from 2 to 7 kHz, helping to locate the sound. This is modeled through the

Results

In [4] the computational model of the peripheral auditory system was presented, with the results that validate the using of the speech processor proposed. However, in order to evaluate the effectiveness of this speech processor, some information from each frequency band is obtained by the neural spike patterns computed both on the corresponding levels of sound signal amplitude and the frequency of the constituent signals.

Fig. 4 shows the neurotransmitter concentrations between IHC and ANF

Discussion

The cochlear stimulation proposed has the advantage of simulating a peripheral auditory system with characteristics more natural than those that produce available implants. In natural hearing, neurotransmitters are constantly released in the IHC synapses, even without auditory stimuli. Spontaneously or stochastically, neurons generate impulses toward the CNS. Some of the current strategies have included the so-called conditioning impulses in order to simulate those spontaneous processes [1],

Conclusions

The present study tries to develop and simulate a cochlear stimulation variant, which is more similar to the physiological characteristic of the human peripheral auditory system. The main advantages of this stimulation variant are the following: a more natural stimulation will produce better understanding by an improved vowel and consonant recognition; the reduction of the amount of bits that the speech processor sends to the implant will lead to a considerable decrease of bandwidth in the

Ernesto Arturo Martínez-Rams was born in Santiago de Cuba, in 1967. He obtained Electronic Engineering degree in 1990 and M.S. in “Telecommunication Systems”, in 2000. He is teaching and research since 1992. Belong at the Telecommunications Department in the University of Oriente, Cuba. Also he has been Head of Electrotechnology Department from 2001 to 2003. He has wide university teaching experience in subjects of electronics and telecommunications technology, building the teaching contents in

References (28)

  • I.C. Bruce et al.

    The effects of stochastic neural activity in a model predicting intensity perception with cochlear implants: low-rate stimulation

    IEEE Transactions on Biomedical Engineering

    (1999)
  • I.C. Bruce, Spatiotemporal coding of sound in the auditory nerve for cochlear implants, Thesis, Department of...
  • P.C. Loizou, Signal processing for cochlear prosthesis: a tutorial review, in: Proceedings of the 40th Midwest...
  • J.P. Martens, L. Van Immerseel, An auditory model based on the analysis of envelope patterns, in: International...

    • Cited by (4)

    Ernesto Arturo Martínez-Rams was born in Santiago de Cuba, in 1967. He obtained Electronic Engineering degree in 1990 and M.S. in “Telecommunication Systems”, in 2000. He is teaching and research since 1992. Belong at the Telecommunications Department in the University of Oriente, Cuba. Also he has been Head of Electrotechnology Department from 2001 to 2003. He has wide university teaching experience in subjects of electronics and telecommunications technology, building the teaching contents in most cases. He is co-author of technical paper and contributions in the fields of electronic design and simulation, audio signal processing and cochlear implant. He has been participated as presenter, reviewer and chairman in international events and symposiums.

    Vicente Garcerán Hernández was born in Cartagena, Spain, in 1956. He obtained Industrial Engineer degree in 1984, M.S. of “Design, Robotics and y Industrial Automation” in 1994, and doctor's degree in 2000. He is Adjunct Professor of Electronics from 1990 to 2002, Full Professor since 2002 and he is Principal of Electronics and Computer Technology Department, in the Technical University of Cartagena, Spain since 2004. He has wide university teaching experience in subjects of electronics and computer technology. He is co-author of technical paper and contributions in the fields of electronic design, audio signal processing, artificial neural networks, speaker recognition and bioinspired systems.

    J. Manuel Ferrández Vicente was born in Elche, Spain in 1969. He received the M.Sc. degree in Computer Science in 1995, and the Ph.D. degree in 1998, all of them from the Universidad Politécnica de Madrid, Spain. He is currently Associate Professor at the Department of Electronics, Computer Technology and Projects at the Universidad Politécnica de Cartagena and Head of the Electronic Design and Signal Processing Research Group at the same University. His research interests include bioinspired processing, neuromorphic engineering and auditory processing.

    View full text
    Copyright © 2008 Elsevier B.V. All rights reserved.