WO2005097257A1

WO2005097257A1 - Aperiodic stochastic resonance applied to cochlear implants and sonar array

Info

Publication number: WO2005097257A1
Application number: PCT/GB2005/001323
Authority: WO
Inventors: Robert Morse; Nigel Geoffrey Stocks
Original assignee: University Of Warwick
Priority date: 2004-04-06
Filing date: 2005-04-05
Publication date: 2005-10-20
Also published as: GB0407803D0

Abstract

A method and apparatus are disclosed for signal dependent noise coding an input signal. As a result of the signal dependent noise coding the rate at which signal information can be transmitted is improved. Aperiodic Stochastic Resonance is applied to cochlear implants and sonar array. A noise-modulated signal is created by multiplying the input signal by the noise component.

Description

APERIODIC STOCHASTIC RESONANCE APPLIED TO COCHLEAR IMPLANTS AND SONAR ARRAY

The present invention relates to a method and apparatus for transmitting signal information at a increased rate. In particular, but not exclusively, the present invention relates to a method and apparatus for applying signal dependent noise coding to an input signal whereby the rate at which signal information can be transmitted is increased over known transmitting systems.

Problems associated with the transmission of information are well known. For example the twin disciplines of information theory and coding theory have been developed with a basic goal being the efficient and reliable communication of information in an imperfect environment. It is well known that to be efficient transfer of information should occur without prohibitively large amounts of time and effort. On the other hand to be reliable it should be possible to reassemble a received data stream to reproduce the transmitted data stream.

As a basic for communication information passes from a source to a sink via a channel. Known encoding schemes determine exactly how information is structured at a source and via a decoder, how information is handled at the sink. The behaviour of the channel of communication is generally not known. There are many forms of environmental factors which may conspire to make the channel less than perfect. As the imperfections caused by an uncooperative or possibly hostile environment increase it becomes increasingly difficult to accurately reassemble the transmitted data stream from a received data stream. It is now widely established that, via an effect termed stochastic resonance (SR) , the addition of random noise to an information bearing signal can lead to an improvement in the amount of information transmitted. This is particularly so when the information is transmitted through a non-linear system. Furthermore it has also been established that another form of stochastic resonance termed suprathreshold stochastic resonance (SSR) can occur in parallel arrays comprised of many non- linear elements. SR and SSR may occur when the constituent elements are non-linear by virtue of an inherent "threshold". Effectively the threshold which is set in a non-linear device determines whether the output from the non-linear element is high or low if the input signal to the non-linear element is above or below a predetermined threshold respectively. SSR has the advantage over SR in that it leads to higher transmission information rates and can occur for a wider range of signal types and characteristics. Nevertheless problems with such known techniques are known and attempts are continually made to try to establish techniques for increasing a rate of information transfer from a source to a sink.

It is an aim of embodiments of the present invention to at least partly mitigate the above-referenced problems.

It is an aim of embodiments of the present invention to provide a method and apparatus for improving the rate at which signal information can be transmitted. Preferably the transmittal occurs through a non-linear array.

It is a further aim of embodiments of the present invention to provide a method and apparatus for processing an input signal so that high transmission information rates can be achieved over a wide range of signal types and characteristics .

Embodiments of the present invention may be adapted for many different applications. Where information is transmitted through a non-linear system. One such application mentioned by way of example is in the field of cochlear implants.

In the normal acoustic system of a human or animal, sounds received at the outer ear are converted into electrical signals by hair cells within the inner ear or cochlea. The electrical signals are subsequently conveyed to the cochlear nerve (the nerve of hearing) to the higher auditory centres of the brain. The hair cells within the cochlea are physiologically vulnerable and can be damaged or completely destroyed by some congenital conditions. Possible destructive conditions are some diseases (such as meningitis), the side effects of some life-saving drugs (such as aminoglycoside antibiotics) or by other causes. The severe loss of hair cells results in profound deafness. Fortunately the cochlear nerve is largely unaffected by the conditions that damage the hair cells and the hearing of profoundly deaf people can often be partially restored by direct electrical stimulation of the cochlear nerve by a cochlear implant.

Many forms of cochlear implants are known. A standard cochlear implant contains three parts: an array of typically 16 to 22 electrodes surgically implanted within the cochlea, a speech processor outside the body for converting the signal at a microphone into the appropriate electrode signals and a radio-frequency link or hardwire link for transmitting the speech processor signals across the skull to the electrodes.

The purpose of the multiple electrodes inside the cochlea is to mimic, albeit crudely, the coding of frequency that occurs with normal acoustic stimulation. In the normal ear, the frequency of an acoustic signal can potentially be coded both in terms of which the fibres are most active (place coding) and the temporal pattern of activity of each fibre (time coding) . The place coding of information is mimicked by filtering the input signal by a bank of band pass filters each with a different pass band. In accordance with the normal tonotopic arrangement of the cochlea, band pass filters with low- frequency pass bands are used to stimulate electrodes in the apex of the cochlea and those with a high-frequency pass band are used to stimulate electrodes in the base.

Known cochlear implants are however imperfect as the signal processing which occurs prior to stimulation of the nerve fibres relies on processed signals which may incorrectly stimulate cochlear nerves resulting in a user hearing either a distorted signal or an inaccurate signal .

It is an aim of embodiments of the present invention to at least partly mitigate the problems above-mentioned in respect of cochlear implants.

It is an aim of embodiments of the present invention to provide an improved cochlear implant . The improved cochlear implant may enhance the speech cues contained in a temporal pattern of nerve activity. As noted above embodiments of the present invention may be adapted for many different applications. Another application mentioned by way of example is in the field of sonar arrays.

Echo sounding has been used for many years as a technique for measuring water depths and detecting objects under water. Sonar detectors can use "active sonar" i.e. for the detection of previously transmitted narrow band pulses or "passive sonar". Passive sonar is effectively a listening array which include a number of hydrophones for listening to underwater objects. In passive sonar no emitted beam is necessary. The hydrophones are broadband receivers rather than the narrow band receivers used in active sonar. A common way of implementing these arrays is to connect each hydrophone to a comparator (i.e. a threshold device) which acts as a polarity detector. These non-linear devices can then be used to extract directional information. However known sonar arrays are limited in their accuracy and in the detail which can be observed.

It is aim of embodiments of the present invention to at least partly mitigate the problems above-mentioned in respect of sonar arrays.

According to a first aspect of the present invention there is provided a method of signal dependent noise (SDN) coding an input signal to thereby increase the rate at which signal information can be transmitted.

According to a second aspect of the present invention there is provided an apparatus arranged to provide signal dependent noise (SDN) coding of an input signal to thereby increase the rate at which signal information can be transmitted.

According to a third aspect of the present invention there is provided a cochlear implant for improving speech comprehension in a target user comprising: at least one electrode implantable in the target user; a microphone for receiving sounds and generating electrical signals responsive to said sounds; a signal processing unit for processing the generated electrical signals and providing at least one output signal; and a transmission link for transferring said at least one output signal to said at least one electrode; wherein said signal processing unit includes at least one coding unit for providing a signal dependent noise coded input signal .

According to a fourth aspect of the present invention there is provided a sonar array for detecting underwater objects comprising: one or more hydrophones at least one of which is arranged to provide an input signal to respective signal processing apparatus arranged to provide signal dependent noise coding of the input signal.

Embodiments of the present invention provide a method and apparatus for coding an input signal which improves the rate at which signal information can be transmitted relative to the rate at which signal information can be transmitted with prior art signal processing technologies. Preferably an input signal is coded to include a signal dependent noise (SDN) component and the transmittal occurs in at least one non-linear element. Embodiments of the present invention provide an improved cochlear implant in which a signal dependent noise component is added to an input signal and the thus SDN coded signal is transmitted to electrodes of a cochlear implant .

Embodiments of the present invention provide an improved signal processing technique for use in sonar arrays. Signal dependent noise is generated in multiple coding units each of which receives a respective input from a respective hydrophone. The thus SDN coded signal is transmitted to comparators which comprise non-linear devices and as a result an improved resolution and accuracy of sonar arrays may be achieved.

Embodiments of the present invention will now be described hereinafter by way of example only with reference to the accompanying drawings in which:

Figure 1 illustrates prior art coding units;

Figure 2 illustrates an embodiment of a coding unit in accordance with the present invention;

Figure 3 illustrates a coding unit in accordance with a further embodiment of the present invention;

Figure 4 illustrates a comparison of information rates;

Figure 5 illustrates a further embodiment of a coding unit; Figure 6 illustrates a further embodiment of a coding unit;

Figure 7 illustrates a schematic of a cochlear implant;

Figure 8 illustrates a cochlear implant;

Figure 9 illustrates one channel of a cochlear implant according to an embodiment of the present invention; and

Figure 10 illustrates an implementation of the present invention in a sonar array.

In the drawings like reference numerals refer to like parts .

Throughout this specification the terra "information" will be referred to. This term is to be given the quantitative definition used in information theory. Quantitatively, the amount of Shannon information I(x,y) that a signal y (termed the output or received signal) conveys about a signal x (termed the input or transmitted signal) is defined as:

I

Where P(x,y) is termed the joint probability density between x and y and P(x), P(y) are the marginal distributions. This quantity is also referred to as the mutual information or the transinformation . As noted above it is now widely established that via an effect termed stochastic resonance (SR) the addition of random noise to an information bearing signal can lead to an improvement in the amount of information transmitted through a non-linear system. A non-linear system is any system or device or element in which the output does not vary linearly with an input. An artificial neuron or biological neuron are examples of non-linear elements . A general example is a threshold device in which a non- linear element is allocated a threshold value. When an input signal is below a threshold value an output from the non-linear element takes a first value or range of values . When the input signal reaches or exceeds the predetermined threshold value the output of the non- linear element takes a second value or second range of values. Stochastic resonance occurs when the presence of noise in a non-linear system can induce an optimal (or improved) output from that system. Another form of stochastic resonance termed suprathreshold stochastic resonance (SSR) can occur in parallel arrays comprised of many non-linear elements. SSR has an advantage over SR in that it leads to higher transmission information rates and can occur for a wider range of signal types and characteristics .

Stochastic resonance can thus loosely be defined as occurring when an increase in noise leads to an increase in output signal to noise ratio in a non-linear system. Examples of non-linear systems in which stochastic resonance has been observed are electronic devices, ring lasers, super conducting quantum interference devices (SQUIDs) and in biological sensory neurons. In such neurons it is an essential qualitative feature that the output of a neuron produces an action potential (spike) when the input signal increases above a threshold. Following the spike there is a period during which a neuron may not spike again. Thus for stochastic resonance in threshold systems such as a biological or artificial neuron when noise is added to an input signal that is too small to cause the neuron to spike alone without the introduction of noise, an output spike can occur which is correlated with the amplitude of the input signal. Effectively adding random noise is equivalent to adding random changes in the values of a neurons threshold. As a result an originally sub-threshold signal occasionally may become suprathreshold.

Suprathreshold stochastic resonance occurs in an array of threshold devices which are the subject of an identical input signal but which have multiple independent noise sources. The output from each device in the array of devices may then be summed to give an overall output.

Figure 1 illustrates a system 10 comprising a parallel array of N coding units 11 and N non-linear elements 12.

A common signal S is fed into the N coding units. Each of the coding units has an independent noise source which is added to the signal. The output from each coding unit is then input into a respective non-linear device 12 and then the outputs are summed via a summer 13. The output from the summer 13 is the array output Y.

The addition of noise leads to an improvement in the transmitted information I(Y,s) between the input signal S and the array output Y={Yι, Y₂, ...Y_N} .

A further improvement in the transmitted information I(Y,s) can be achieved in accordance with the present invention. Rather than providing a noise source independent of an input signal, as proposed in the prior art, the two quantities are coupled. In this way the noise intensity is derived from the signal intensity and thus gives rise to signal-dependent-noise (SDN) coding. SDN coding leads to an enhancement of the SR and SSR effect and thus to an enhancement of I(Y,s) .

Figure 2 illustrates a first embodiment of the present invention in which SDN coding can be achieved. As a general point however the criterion for SDN coding is that at some point the signal and noise must be multiplied (or a function that provides multiplicative components may be used) . Alternatively the integrated effect of signal and noise at a non-linear device may be multiplicative.

Figure 2 illustrates a coding unit 20 which receives a signal S at input 21. This signal is half wave rectified via rectifier 22 and then filtered with a suitable filter 23. It will be understood that both of these stages are optional. The rectified and filtered signal is then scaled and the DC component adjusted in a scaling and offset adjustment unit 24. It will be understood that if scaling and offset adjustment is not required this stage may be omitted. The resultant signal is then multiplied via multiplier 25 with a noise signal generated from an independent noise source 26 which (if necessary) has itself been suitably scaled and has had its DC component adjusted by a scaling and offset adjustment unit 27. The output coded signal 28 is a signal dependent noise coded form of the input signal. This can be input into a nonlinear device and the rate at which signal information can be transmitted is increased. Figure 2 thus illustrates a technique in which at least one signal dependent noise signal can be generated responsive to an input signal and which provides at least one signal dependent noise signal as an output signal.

Figure 3 illustrates a further example of a coding unit 30 according to a second embodiment of the present invention in which an input signal input at node 31 is input into a half wave rectification block 32 and a filter 33. It will be understood that the rectification and filtering steps are optional. Signal scaling and offset adjustment takes place at block 34 (this is also an optional step) and then the output of this is multiplied in multiplier 35 with the output from an independent noise source 36. This noise source may be optional in which case noise generated internally rather than artificially may be used and the output of the independent noise source 36 may itself be scaled and offset adjusted prior to input at the multiplier 35. The output from the multiplier 35 is a generated signal dependent noise signal which is responsive to the input signal at 31. This output signal is summed with the signal input at 31 in a summer 37. The resultant summed signal forms an output signal 38 which can be input into a non-linear device.

As will be noted by those skilled in the art in accordance with the first embodiment of the present invention no summing unit is required in the coding unit 20. By contrast in accordance with the second embodiment a summer is used but an amount of noise that depends multiplicatively on the signal intensity is added to the signal . In accordance with the first embodiment of the present invention the output of the coding unit 20 Y can be written as: y x (f (s) + a) (n+β)

Where f (s) is the processed signal (possibly half or full wave rectified and filtered) and α and β are the signal and noise DC levels respectively. By adjusting the scaling (noise and signal) and the offsets (α and β) various SDN coding strategies can be implemented. For example taking α=β=0 leads to Y <x f (s) n . Here the processed signal modulates (with an effective modulation depth of 100%) a noisy carrier.

Different modulation depths may be implemented by selecting values of the parameters β and α. Nonzero α and β leads to a modulated carried having an arbitrary modulation depth plus an additional component f (s) β. If a,β»\ then a predominantly additive strategy where noise is only weakly dependent upon the signal can be implemented. It will be understood that the dependence of the noise on the signal can thus be selected.

Both of the coding units illustrated schematically in figures 2 and 3 describe how an input signal can be signal dependent noise coded. As a result a noise component which is responsive to the input signal is introduced to that input signal. This may be achieved either by multiplying an input signal (or a signal closely dependent to it) with a noise source or by further summing such a multiplied signal with the input signal itself. Figure 4 illustrates a comparison of the information rate achieved using SDN coding with a conventional additive strategy. The right panel indicated as figure 4b shows an enhanced SR and SSR effect compared with the additive coding scheme illustrated in figure 4a. Both figures 4a and 4b indicate information rates through a system including at least one coding unit and at least one nonlinear array. Both figures 4a and 4b show the results when non-linear elements including one fibre or including sixteen fibres are used. The fibres refers to the type of non-linear elements used in the study (integrate and fire neurons) . In this way the fibres relate to the number of inputs into a final decision making unit.

Figure 5 illustrates a coding unit 50 according to a further embodiment of the present invention. The coding unit 50 receives a signal S at input 51. This signal which is the original signal is input into a summing unit 52 and a further noise source 53 is used to generate a further input to the summing unit 52. The embodiment described with respect to figure 5 thus provides a modification to the embodiment disclosed in figure 2. A further noise source is added. The signal output from the summing unit 52 is half wave rectified via rectifier 54. The output from this unit is then filtered via a suitable filter 55. It will be understood that both the rectification and filter stages are optional in this embodiment. The rectified and filtered signal is then scaled and the dc component adjusted in a scaling and offset adjustment unit 56. It will be understood that if the scaling and offset adjustment is not required this stage may also be omitted. The resultant signal is then multiplied via a multiplier 57 with a noise signal generated from an independent noise source 58 which (if necessary) has itself been suitably scaled and has had its dc component adjusted by a scaling and offset adjustment unit 59. The output coded signal 60 is a signal dependent noise coded form of the input signal. This can be input into a non-linear device and the rate at which signal information can be transmitted is increased compared to prior art techniques . Figure 5 thus illustrates a technique in which at least one signal dependent noise signal can be generated responsive to an input signal and which provides at least one signal dependent noise signal as an output signal.

Figure 6 illustrates a still further coding unit according to a further embodiment of the present invention. The coding unit 61 includes a signal input at node 62 which forms an input into a summing unit 63. A noise source 64 generates noise which also forms an input into the summing unit 63. The remaining elements of the coding unit are similar to those described with respect to figure 3. In this respect the output from the summing unit 63 is input into a half wave rectification block 65 and the output from this into a filter 66. It will be understood that the rectification and filtering steps are optional. Scaling and offset adjustment take place in a signal scaling and compression unit 67 (this is also an optional step) and then the output from this adjustment unit is multiplied in multiplier 68 with the output from an independent noise source 69. It will be understood that noise scaling and offset adjustment of the output from the noise source may be used if required. The noise source 69 may be an independent noise source or optionally may be noise generated internally rather than artificially. In this sense the indication of a separate noise source is for schematic purposes only. The output from the multiplier 68 and summer 63 form inputs to a further summation block 70. The resultant summed signal 71 forms an output signal which can be input into a non- linear device.

As will be appreciated by those skilled in the art in accordance with the embodiment described with respect to figure 5 a single summing unit is required in the coding unit. By contrast in accordance with the embodiment described with respect to figure 6 two summing units are used.

It has been appreciated that the addition of an independent noise source to the signal prior to entry into a coding unit as described with respect to figures 2 and 3 leads to an improved information transmission in some parameter ranges .

Embodiments of the present invention thus provide a method and apparatus for signal processing in which an input signal is coded with signal dependent noise. As a result the rate of flow of information through the coding unit and a respective non-linear device, such as an artificial neuron, can be increased compared to conventional information rates. Any appropriate method may be used for introducing signal dependent noise. The dependency of the noise in the signal dependent noise signal may be controlled by varying predetermined parameters.

As noted above embodiments of the present invention can be applied to many different types of technologies. Figure 7 illustrates a cochlear implant. As noted above hair cells within the cochlea in a human inner ear are physiologically vulnerable and can be damaged or completely destroyed. The severe loss of hair cells results in profound deafness. It has been found that the hearing of profoundly deaf people can often be partially restored by direct electrical stimulation of the cochlear nerve (which remains substantially undamaged) by a cochlear implant. The cochlear implant 70 contains a number of features. These are a microphone for picking up sounds, a signal processor for converting the sound into electrical signals, a transmission system of transmitting the electrical signals to implanted electrodes and an electrode or electrode array which is inserted/implanted into the cochlea of a target individual by a surgeon. As illustrated in figure 7 the microphone 71 picks up the sounds which may include a speech signal or non speech signal. A signal processor box worn by a patient processes the sound signal from the microphone with a pre-processing filter 72 and then multiple pass band filters 73. These divide the acoustic wave form into a number of channels (three are shown by way of example) . It will be understood that more or less of the pass band filters 73 may be used. The number of channels is proportional with the number of electrodes implanted in the target patient. The output of each respective pass band filter is input into a respective post processing block 74 and are then transmitted to the electrodes implanted in a user. In figure 7 the transmission is illustrated as a radio frequency link requiring a percutaneous plug. It will be understood that the signals from the post processing blocks may alternatively be transmitted via a hardwired link. In this way the relative amplitudes of the current pulses delivered to the electrodes are proportional to the spectral content of an input signal.

The purpose of the multiple electrodes located inside the cochlea is to mimic, albeit crudely, the coding of frequency that occurs with normal acoustic stimulation. In the normal ear the frequency of an acoustic signal can potentially be coded both in terms of which nerve fibres are most active (place coding) and the temporal pattern of activity of each nerve fibre (time coding) . The place coding of information is mimicked by filtering the input signal by a bank of band pass filters each with a different pass band. In accordance with normal tonotopic arrangements, the cochlear band pass filters with low- frequency pass bands are used to stimulate electrodes at the apex of the cochlea and those with a high-frequency pass band are used to stimulate electrodes in a base region. The band pass filtering of the input is part of what is generally termed a speech coding strategy which in modern prior art implants may be implemented by software in the external speech processor. In this way it can be changed without changes being made to the hardware. Although the terms "speech processor" and "speech coding strategies" are generally used the cochlear implant input is also intended to process non- speech input (environmental sounds) such as background traffic, doorbells and so on. A particular cochlear implant can have more than one strategy in software and the user can switch between them. The various strategies can be either just slight changes in the parameters (used for example in poor listening conditions) or a completely different strategy. For example certain implants can implement simultaneous analogue stimulation (SAS) , continuous interleaved sampling (CIS) and/or paired pulsatile sampler (PPS) .

Ideally a speech coding strategy will evoke the same patters of nerve activity with electrical stimulation as those that are evoked in the normal ear with acoustic stimulation. However because of the reduced number of healthy nerve fibres present in a damaged cochlea and the limited number of electrodes that can be surgically implanted into the cochlea, improvements are limited. Typically 20 or so electrodes in the cochlea will stimulate about 15000 nerve fibres. Furthermore the electrodes are surrounded by normal conductive fluids of the cochlea and the currents from the individual electrodes therefore tend to spread throughout the conductive medium. Consequently nerve fibres are not under the control of single electrodes as desired. However given that the cochlear nerve is the only neural pathway between the ear and higher auditory centers it is most helpful that essential auditory cues, particularly speech cues are represented by the space and temporal activity of the cochlear nerve level. If information is not represented at this level then it cannot be reconstructed at higher levels .

All implant strategies map the large intensity range of natural acoustic signals (about lOOdB from threshold to discomfort) onto the narrow dynamic range for electrical stimulation of the cochlear nerve (typically lOdB) . This is achieved by instantaneous compression of the cochlear implant signals which is part of a post-processing process carried out by the post processing units 74. In many known speech coding strategies the electrode currents are trains of biphasic pulses that are temporarily interleaved. The designers of the known strategies have assumed that high frequency temporal information is not usable by implantees and the instantaneous amplitude of the pulses is derived from the envelope of the band-pass filter output. In other strategies auditory cues are intended to be coded by both time and place cues. The compressed outputs of the filters are used directly to stimulate segments of the cochlear nerve and the high frequency temporal information in each channel output is presumed to be retained in the pattern of evoked nerve discharges. Given the non-linear "all-or-nothing" characteristic of the nerve such presumption is not justified. Strategies in which the fine-time structure of the stimulus are transmitted to the electrodes rather than envelope information are referred to as analogue strategies even when they are implemented in software .

According to embodiments of the present invention the addition of random signals (noise) to cochlear implant signals may lead to improved performance by enhancing the speech cues contained in the temporal pattern of nerve activity. The noise may induce the nerve fibres to be spontaneously active in the absence of an auditory input. Spontaneous neural activity is known to occur in normal hearing but has been largely absent in patients fitted with cochlear implants in the past. It has been demonstrated by the present inventors that via the suprathreshold stochastic resonance effect the information transmitted by a neural population is improved if independent noise sources are added to nerve fibres. The noise leads to a partial decorrelation of neural activity across the population that results in an improvement of the information transmitted.

Embodiments of the present invention provide a method and apparatus in which the amplitude of noise introduced into a cochlear implant should be related to the amplitude of the signal. This is achieved by multiplying the noise wave form (after any appropriate scaling and offset adjustment) by the information bearing signal.

Figure 8 illustrates a cochlear implant according to an embodiment of the present invention. The cochlear implant 80 includes a microphone 81 which is arranged to pick up sounds which may include voices and to convey an analogue signal responsive to those received sounds to a signal processor pack 82 carried by a patient. The processor pack 82 includes signal processing software and/or hardware which receives the analogue signals from the microphone and outputs a signal via radio transmitter 83 to one or more electrodes 85 via a radio receiver implanted into a target user such as a patient. Alternatively the output from the processor pack may be hardwired to a transmitter attached to a user so that the radio link between a transmitter and receiver need only be made over a short path. The connection may alternatively be via a hardwire connection as is known by those skilled in the art.

As noted in figure 8 a plurality of signal processing chains are provided. One each for the number of electrodes implanted in a patient. The signal from the microphone 81 forms a common input at node 86 and this input is then independently processed via signal processing units 87. The output from each signal processing unit is connected to the multi core wire 83 so that an output from each signal processing unit may independently be connected to a respective electrode. As an alternative the signals may be multiplexed.

Figure 9 illustrates one channel of the signal processing carried out in the pack 82 of a cochlear implant using signal-dependent noise coding according to an embodiment of the present invention. As such figure 9 illustrates just one possible way in which SDN coding can be applied to a cochlear implant using an analogue strategy. It will be understood that there would be an appropriate channel for each electrode in the cochlear implant. According to this embodiment rather than stimulating the electrode with the compressed output of a band pass filter directly, as would be the case with prior art cochlear implants, the output is first coded to introduce a signal dependent noise element. This may be achieved by multiplying the input signal by a gaussian noise waveform. It will be understood that other analogue noise waveforms other than gaussian noise sources may be used. Indeed the spectral density and probability distribution of the noise waveform are characteristics that may be varied to optimise performance. Also pulsatile carriers similar to those currently used in SPEAK, ACE and CIS strategies may be employed. In such implementations the output from the multiplier (i.e. the output from the SDN coding unit) would be used to modulate a characteristic of the pulse train such as the pulse amplitude, pulse duration and/or inter-pulse interval .

As illustrated in figure 9 the analogue input 90 from the microphone input after AGC and pre-processing is input into a band pass filter 91. The output from the band pass filter forms an input into a compression unit 92 and the output of this forms the input to an SDN coding unit 93. Either of the coding unit examples shown in figures 2 and 3 and 5 and 6 may be used for the coding unit 93 with one or more appropriate noise source 95. The output of the coding unit is input into a post processing block 94 and then output via a fixed link or radio-transmission link 83 to the implanted electrodes.

It will be understood that by coding the analogue input signal from the microphone with signal dependent noise and by stimulating an electrode connected to the cochlear nerve in a user the comprehension of a user to sounds heard may be increased.

Figure 10 illustrates how signal dependent noise coding can be applied to sonar arrays in accordance with a still further embodiment of the present invention. Sonar arrays are underwater listening devices that can be used for a wide variety of applications. The present invention is most suited to passive sonar arrays. These arrays are predominantly used in anti-submarine applications for detecting mainly submarines but also other forms of shipping. The arrays of hydrophones have two main functions, firstly to detect whether a signal (target) is present and second to classify the signal type. This second task requires that sufficient information is transmitted about a signal source. It will be understood that other embodiments of the present invention may improve other broadband sonar applications such as underwater communication. The sonar array includes one or more hydrophones 100ι, 100₂... 100_N which detect sounds under the water. The sonar arrays may be used in anti-submarine applications. The output from each respective hydrophone is input into a respective coding unit 101 at which the signal from the hydrophone is signal dependent noise coded. This introduces a noise component which is dependent upon the input signal to the signal as explained above and in particular as shown in figures 2 and 3 and 5 and 6. The coded output signal is input into a respective comparator 102 which compares the signal output from a respective coding unit with a predetermined reference voltage. Each comparator outputs a respective output signal Y which may be used when assembling the sonar image.

Embodiments of the present invention are generally applicable to coding input signals prior to introduction into a non-linear system to thereby increase the rate of information flow. Embodiments of the present invention may therefore be applied to electronic devices, ring lasers, SQUIDs and other devices.

Embodiments of the present invention may also be applicable to electronic stimulation of further parts of the auditary processing system (for example the cochlear nucleus) .

Throughout the description and claims of this specification, the words "comprise" and "contain" and variations of the words, for example "comprising" and comprises", means "including but not limited to", and is not intended to (and does not) exclude other moieties, additives, components, integers or steps.

Throughout the description and claims of this specification, the singular encompasses the plural unless the context otherwise requires. In particular, where the indefinite article is used, the specification is to be understood as contemplating plurality as well as singularity, unless the context requires otherwise.

Features, integers, characteristics, compounds, chemical moieties or groups described in conjunction . with a particular aspect, embodiment or example of the invention are to be understood to be applicable to any other aspect, embodiment or example described herein unless incompatible therewith.

Embodiments of the present invention have been described hereinabove by way of example only. It will be understood that modifications may be made to the specific examples without departing from the scope of the present invention .

Claims

CLAIMS :

1. A method of signal dependent noise (SDN) coding an input signal to thereby increase the rate at which signal information can be transmitted.

2. The method as claimed in claim 1 further comprising: subsequent to SDN coding the input signal, inputting the coded signal into at least one non-linear device.

3. The method as claimed in claim 1 or 2 further comprising the steps of: generating at least one signal dependent noise (SDN) signal responsive to said input signal; and providing at least one output signal responsive to each respective SDN signal.

4. The method as claimed in claim 3 further comprising the steps of: providing at least one SDN signal as a respective output signal .

5. The method as claimed in claim 3 further comprising the steps of: summing said input signal with said SDN signal; and providing the resulting signal of said summation as said output signal.

6. The method as claimed in claim 3 wherein said step of generating each SDN signal comprises: multiplying a respective random or pseudo-random noise signal with said input signal or a signal derived therefrom.

7. The method as claimed in claim 1 wherein said step of generating each said SDN signal comprises: applying a function that produces multiplicative components to said input signal and a respective noise signal.

8. The method as claimed in claim 6 or claim 7 wherein each respective random noise signal is independent of other random noise signals used to generate other SDN signals.

9. The method as claimed in any one of claims 1 to 8 further comprising the steps of: prior to said step of generating an SDN signal, rectifying and/or filtering said input signal.

10. The method as claimed in any one of claims 1 to 9 further comprising the steps of: for each SDN signal generated, providing a random noise source and carrying out a scaling and offset adjustment operation on a random noise signal provided by said random noise source prior to said step of generating the SDN signal.

11. The method as claimed in any one of claims 1 to 10 further comprising the steps of: providing each said output signal as an input to a respective non-linear threshold device.

12. The method as claimed in claim 11 further comprising the steps of: summing together an output from each respective threshold device.

13. Apparatus arranged to provide signal dependent noise (SDN) coding of an input signal to thereby increase the rate at which signal information can be transmitted.

14. The apparatus as claimed in claim 13 further comprising: at least one non-linear device arranged to receive an SDN coded signal.

15. The apparatus as claimed in claim 13 or 14 further comprising : at least one coding unit each arranged to receive a common input signal and comprising: a signal generator for generating a respective signal dependent noise (SDN) signal responsive to said common input signal; wherein said coding unit is arranged to provide a coded output signal responsive to said SDN signal.

16. The apparatus as claimed in claim 15 wherein said coded output signal of each coding unit comprises the SDN signal generated by the coding unit.

17. The apparatus as claimed in claim 15 wherein each coding unit further comprises a summer device and said coded output signal comprises the sum of the SDN signal generated by the coding unit, and said common input signal .

18. The apparatus as claimed in any one of claims 15 to 17 wherein each coding unit further comprises an independent random noise source.

19. The apparatus as claimed in any one of claims 15 to 18 wherein said apparatus further comprises: at least one non-linear threshold device, each associated with a respective one coding unit; wherein the coded output signal from the respective coding unit provides an input signal to said threshold device.

20. The apparatus as claimed in claim 19 further comprising : a summer device for summing an output signal from each threshold device and providing a summer output signal indicative of a characteristic of said common input signal.

21. A cochlear implant for improving speech comprehension in a target user comprising: at least one electrode implantable in the target user; a microphone for receiving sounds and generating electrical signals responsive to said sounds; a signal processing unit for processing the generated electrical signals and providing at least one output signal; and a transmission link for transferring said at least one output signal to said at least one electrode; wherein said signal processing unit includes at least one coding unit for providing a signal dependent noise coded input signal .

22. The cochlear implant as claimed in claim 21 wherein each coding unit comprises: an input for receiving an input signal; and a signal generator for generating at least one signal dependent noise (SDN) signal responsive to both said input signal and a noise signal provided from a noise source.

23. The cochlear implant as claimed in claim 22 wherein said signal generator comprises a multiplier for multiplying the input signal or a signal derived therefrom with the noise signal.

24. The cochlear implant as claimed in any one of claims 21 to 23 wherein each said coding unit is arranged to provide the generated SDN signal as a coded output signal .

25. The cochlear implant as claimed in any one of claims 21 to 23 wherein each said coding unit further comprises a summer device for summing two input signals, and is arranged to provide the sum of the generated SDN signal and the input signal as a coded output signal.

26. A sonar array for detecting underwater objects comprising : one or more hydrophones at least one of which is arranged to provide an input signal to respective signal processing apparatus arranged to provide signal dependent noise coding of the input signal.

27. The sonar array as claimed in claim 26 further comprising : for each signal processing apparatus, a comparator for comparing the signal dependent noise coded input signal with a predetermined threshold level and outputting a respective output signal responsive to the result of a comparison made.

28. A method substantially as hereinbefore described with reference to the accompanying drawings.

29. Apparatus constructed and arranges substantially as hereinbefore described with reference to the accompanying drawings .