WO2010000042A1

WO2010000042A1 - Linear gain amplification for mid-to-high intensity sounds in a compressive sound processor

Info

Publication number: WO2010000042A1
Application number: PCT/AU2009/000870
Authority: WO
Inventors: Peter Blamey
Original assignee: Peter Blamey
Priority date: 2008-07-04
Filing date: 2009-07-05
Publication date: 2010-01-07

Abstract

In one embodiment, an apparatus for processing sound includes a means for analysing a sound signal into a number of signal channels that may be classified as desired signal channels and noise signal channels and further divided into frequency specific signal channels; a means for applying variable gain to each signal channel independently; and a means for applying a minimum gain to each signal channel so as to achieve linear amplification of mid-to-high intensity signals in each signal channel. The numerous gain adjusted signal channels are then combined in order to generate a single output sound signal. The application of the minimum gain differentially in each signal channel improves sound quality for loud sounds and improves signal-to-noise ratio in the combined output signal. The apparatus may be implemented in dedicated hardware embodiment or by software running on a microprocessor.

Description

LINEAR GAIN AMPLIFICATION FOR MID-TO-HIGH INTENSITY SOUNDS IN A COMPRESSIVE SOUND PROCESSOR

Field of the Invention: The present invention relates to sound processing devices in which an acoustic sound input or an electric or digital representation of an acoustic sound input is processed and converted to an acoustic sound output, and in particular relates to the processing of sound in the mid to high intensity part of the output sound range to improve speech intelligibility, sound quality and naturalness of the sound. Sound processing devices of this kind are often used in hearing aids, assistive listening devices (ALD), and consumer audio devices such as radios, television sets, CD players, MP3 players, stereo systems, headsets, telephones, and mobile phone handsets. The Global Medical Device Nomenclature Agency (GMDNS) definition of an ALD is an amplifying device, other than a hearing aid, for use by a hard of hearing person.

Background of the invention:

In the field of consumer audio devices, it is common for the sound output level to be controlled by means of a volume control and for the spectral shape to be controlled by a bass or treble adjustment or a multi-channel graphic equalizer. These controls may be adjusted according to the listener's preferences, the input signal to the device, and the ambient sound in the listener's environment. By adjusting these controls, the listener is effectively optimizing the sound output from the device to fall within their preferred listening range in each frequency band taking into account their preferences, their hearing abilities, the sound they are listening to, and the other sounds in the environment. There are also automatic means of achieving some of these desired adjustments, such as automatic gain control (AGC) which responds to changes in the input level (AGCi) or output level (AGCo) of the signal and automatic volume control (AVC) which responds to changes in the level of the ambient sound in the environment. AGCo is also commonly used as a form of hearing protection device or limiter to prevent sounds from becoming loud enough to damage the listener's hearing. AVC is commonly used as a means of keeping the output level above the ambient noise level to maintain audibility, and speech intelligibility and avoid masking of the signal by the noise.

In consumer audio devices, AGC and AVC do not automatically take into account the listener's preferences or hearing abilities. In the fields of hearing aids (HA), cochlear implants (CI), and assistive listening devices (ALD), it is common for parameters derived from the listener's hearing abilities and preferences to be programmed into the device at various frequencies and for AGC, AVC, or other methods to be implemented with the goal of maintaining the sound output of the device within the desired output level range in one or more frequency bands. The primary function of HAs and ALDs is to amplify or process the sounds in the listener's environment, and so the distinction between AGC and AVC is blurred. The input signal is usually equal to the ambient sound in the environment for HAs and ALDs.

In conventional devices for people with normal or impaired hearing, AGC is typically described by an input-output function (for example, see Figure 3), or the relationship between the input level and the gain in the case of AGCi (see Figure 4), or the relationship between the output level and the gain in the case of AGCo (see Figure 5). Figures 3, 4, and 5 are equivalent descriptions of the behaviour of an example AGC device, although the AGCo and AGCi versions would achieve this result with different implementations and methods. It is notable that the high compression regions are characterized by a gain that decreases without limit as the input intensity increases in the case of AGCi, or as the output intensity increases in the case of AGCo. It is also notable that the maximum gain is the gain that applies in the linear region at low to mid intensity levels. In all commonly used non-linear hearing aid fitting prescriptions, such as the NAL-NLl prescription, the FIG6 prescription, the IHAFF prescription, and the DSL i/o prescription (see for example, figure 9.12 of Dillon, H., Hearing aids, Boomerang Press, 2001) a maximum gain is prescribed for a linear region at low intensities and compression is applied at mid-to-high intensities. The conventional rationale for this type of compression is that people with impaired hearing have a steeper than normal relationship between loudness and sound level referred to as recruitment (Fletcher & Munson, 1937). By compressing the signal, the slower growth of the output signal level compensates for the faster growth of loudness. In hearing aids fitted with these prescriptions, the volume control typically controls the maximum gain that can be applied to the signal. In some other types of devices, including CIs and consumer audio devices, the maximum gain is controlled by means of a sensitivity control, and the maximum output level is controlled by means of a volume control.

Compression is also used in AVC systems to ensure that the signal remains in the range between the ambient noise level and the level at which the signal will become uncomfortably loud. As the ambient noise level rises, this target intensity range becomes narrower, and higher compression ratios are required. Compression is usually applied to the high intensity parts of the output signal to keep them below the discomfort threshold of hearing.

The normal human auditory system includes some elements that may be thought of as biological AGC systems. The outer hair cells in the Organ of Corti of the cochlea provide additional gain to the motion of the basilar membrane at low intensities. As the input sound intensity increases, the additional gain provided by the outer hair cells decreases. This is sometimes called "cochlear compression". In contrast to the conventional shapes of input-output curves (as shown in Figure 3), the input output curve for the cochlear compression has its shallowest slope in the middle of the intensity range (see Figure 6). It has been reported that complete loss of outer hair cells can account for up to 60 dB of hearing loss. In addition to the action of the outer hair cells in the cochlea whose function includes adding additional gain to low-intensity sounds, the stapedius muscle and stapedial reflex can act to reduce the intensity of sound transmitted through the mechanical linkage of the middle ear when the sensation becomes too loud. This action of the stapedius muscle is analogous to AGCo, and the action of the outer hair cells is analogous to AGCi. There is also a vocalization- induced stapedius reflex that reduces the sound pressure level reaching the inner ear by about 20 dB when the listener is vocalizing. The vocalization-induced stapedius reflex is analogous to a voice activated change in the input level (a form of AGCi). The stapedius muscle and stapedial reflex are still functional in most people with impaired hearing.

The conventional forms of AGC and AVC use compression in the mid-to-high intensity part of the input and output ranges, in order to compensate for hearing loss or rising ambient noise levels. Compression inevitably introduces distortion into the output signal. The level of the distortion products is proportional to the rate of change of the gain of the device and the input level of the signal. Thus the application of high compression ratios with fast acting compression in the high intensity part of the output range is likely to generate audible distortion products that degrade the quality of the sound. In the case of a person wearing a HA or ALD while listening to speech in loud noise, the compression will reduce the intensity difference between the speech and the noise, resulting in a decrease in intelligibility of the speech. Furthermore, the compression at mid-to-high levels in the HA or ALD is attempting to compensate for a reduction in the outer hair cell compression which naturally occurs at low-to-mid intensity levels. Furthermore, the naturally occurring compression that occurs through the action of the stapedius muscle will operate in addition to the effect of the AGC and/or AVC. Furthermore, unless the device is quite sophisticated, it will be unable to differentiate between the wearer's own vocalization and other speakers' voices so that the effect of the vocalization-induced stapedial reflex and the device compression together will be to artificially reduce the apparent level of the device wearer's own voice relative to other speakers.

Most compression schemes are capable of implementing a linear input-output function throughout their whole range of operation by making the compression ratio 1:1 or by setting the "low kneepoint" to a high input level, however this does not achieve the requirement for the application of compression in the low-to-mid input level range. A few hearing aid compression schemes are capable of implementing compression for low input intensities and a linear or approximately linear input- output function for high input intensities. For example, Goldstein used a model of cochlear compression which tended towards linearity at high input intensities; Armstrong, Sykes, and Csermak used a combination of compression and expansion at high intensities; and Salmi and Scheller summed the outputs of a conventional compression scheme and a linear amplifier. These schemes all provide gains that are close to unity for high input levels, and this is an unnecessary limitation which may result in the output sounds being too soft in the high intensity part of the range. The common fitting prescriptions for hearing aids are not readily adapted to input-output functions with a linear section in the mid-to-high input range, especially if the gain of that section is limited to be close to unity. In addition, the schemes are quite complex in their implementation, involving a model of the cochlear compress ion in the case of Goldstein, a series combination of compression and expansion to achieve the linear section in the case of Armstrong et al, and summation of two parallel compressive and linear amplification paths in the case of Salmi & Scheller. A further shortcoming of these schemes is that they are designed for an in-line signal processing architecture where one or more signal channels are processed in parallel and the outputs from the channels are summed. In-line processing architectures may incur penalties in the form of increased time delay between input and output of the signals, increased power consumption, and increased complexity relative to off-line processing architectures (compare Figures 1 and 2).

Adaptive Dynamic Range Optimization (ADRO®) is an alternative way of overcoming the disadvantages of compression in the mid-to-high intensity part of the dynamic range (Blarney et al). ADRO is an adaptive linear amplification scheme that slowly adjusts gain to compensate for long- term variations in the output signal dynamic range. The peaks in the output signal, represented by the 90^th percentile of the output level distribution for example, are adjusted to fall at or below a "comfort target" and the troughs in the output signal, represented by the 30^th percentile of the output level distribution for example, are adjusted to remain at or above an "audibility target". These two adjustments are referred to as the "comfort rule" and the "audibility rule" respectively. Common implementations of ADRO also use a "hearing protection rule" that applies fast infinite compression at high output levels to prevent the sound becoming uncomfortably loud and potentially damaging the listener's hearing, and a "background noise rule" that imposes a maximum gain to prevent soft background sounds from being over-amplified. ADRO has been shown to produce more natural sound quality than schemes that use compression in the mid-to-high intensity range, and is preferred over compression by approximately 75% of hearing aid users. However, ADRO is reported to make loud sounds too soft in some circumstances. This typically occurs when a loud sound is followed by a soft sound and the slow-acting ADRO rules may take ten seconds or more to increase the gain back to an appropriate level for the second softer sound. Another situation in which loud sounds can be too soft is if they have a narrow dynamic range. If the intensity difference between the 90^th percentile and the 30^th percentile is small relative to the intensity difference between the comfort target and the audibility target, then the gain and output level can be pushed down to very low levels by the comfort rule and the audibility rule will be incapable of raising the gain sufficiently for the peaks to reach the comfort level again.

Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is solely for the purpose of providing a context for the present invention. It is not to be taken as an admission that any or all of these matters form part of the prior art base or were common knowledge in the field relevant to the present invention as it existed before the priority date of each claim of this application.

Throughout this specification the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

Summary of the invention:

According to a first aspect the present invention provides a method for setting and implementing linear gain for mid-to-high intensity signals in a sound processor, the method comprising: providing amplification of the signal with gain under the control of a control module; constraining the gain of the amplifier to remain at or above a minimum gain value for mid-to- high intensity signals; setting, storing, and adjusting the minimum gain value; and calculating one or more appropriate values for the minimum gain value.

According to a second aspect the present invention provides a sound processing device with linear gain for mid-to-high intensity signals, the sound processing device comprising: an amplifier with gain under the control of a control module; a control module that constrains the gain of the amplifier to remain at or above a minimum gain value for mid-to-high intensity signals; a means for setting, storing, and adjusting the minimum gain value; and a means for calculating one or more appropriate values for the minimum gain value.

According to a third aspect the present invention provides a computer program product comprising computer program code means to make a computer execute a sound processing procedure with linear gain for mid-to-high intensity signals, the computer program product comprising: computer program means for providing amplification of the signal with gain under the control of a control module; computer program means for constraining the gain of the amplifier to remain at or above a minimum gain value for mid-to-high intensity signals; computer program means for setting, storing, and adjusting the minimum gain value; and computer program means for calculating one or more appropriate values for the minimum gain value.

The amplifier and control module preferably comprise a conventional AGCi, AGCo, or ADRO sound amplifier with gain control, modified to impose a minimum gain constraint in the mid-to-high intensity region.

The appropriate values for the minimum gain value are preferably calculated on the basis of extraneous factors comprising the hearing loss of the listener, the preferences of the listener, the ambient noise level in the vicinity of the sound processor, the setting of the volume control on the device, and the setting of the sensitivity control of the device.

In a preferred embodiment, the minimum gain constraint may be applied in at least one channel of an in-line signal processor. In another preferred embodiment, the minimum gain constraint may be applied in at least one channel of an off-line signal processor.

Brief description of the drawings:

Figure 1 illustrates a block diagram for a sound processor with off-line processing elements that control an adaptive processor on the signal path. Figure 2 illustrates a block diagram for a sound processor with in-line parallel signal processing paths.

Figure 3 shows an illustrative example of a typical input output function of a sound processor. Figure 4 shows an illustrative example of a typical input gain function of an input controlled automatic gain control (AGCi) in a sound processor. Figure 5 shows an illustrative example of a typical output gain function of an output controlled automatic gain control (AGCo) in a sound processor.

Figure 6 shows a theoretical example of the input output function of the outer hair cells in a normal human ear.

Figure 7 illustrates a generalized channel processor for an embodiment of the present invention with AGCi and minimum gain constraint. Figure 8 illustrates a generalized channel processor for an embodiment of the present invention with AGCo and minimum gain constraint.

Figure 9 shows an illustrative example of a typical input output function of a sound processor with the minimum gain constraint applied in the mid-to-high intensity region. Figure 10 shows an illustrative example of a typical input gain function of an input controlled automatic gain control (AGCi) in a sound processor with the minimum gain constraint applied in the mid-to-high intensity region.

Figure 1 1 shows an illustrative example of a typical output gain function of an output controlled automatic gain control (AGCo) in a sound processor with the minimum gain constraint applied in the mid-to-high intensity region.

Figure 12 illustrates a generalized channel processor for an embodiment of the present invention with ADRO and minimum gain constraint.

Description of the preferred embodiments:

Figure 1 illustrates a system for sound signal processing. One or more input signals 101 are passed to a channel separator 102. The input signals are usually provided by one or more microphones or by signals transmitted from a remote microphone as in a telephone for exampl Iei or from a signal store as in an MP3 player for example. The channel separator 102 provides meaiia iυi optionally separating the input signal(s) into parallel channels for further processing by one or more channel processors 103. In one embodiment of the present invention, the channel separator 102 comprises a bank of bandpass filters and each channel 103 processes the output of one bandpass filter. In another embodiment of the present invention, the channel separator 102 splits the signal into one or more slowly varying parts (the background noise) and one or more rapidly varying parts (the desired signals). Each channel processor 103 processes its input signal to produce one or more control signals 105 that are passed to the adaptive processor 106. The adaptive processor 106 processes the input signals 101 to form one or more output signals 106 from the sound processor. The output signals 106 are usually converted to sound or transmitted to a remote location where they are converted to sound by a loudspeaker or similar transducer. The channel processors 103 may optionally receive or send control signals 104 from or to other channel processors 103. Each channel processor 103 in the embodiment illustrated by Figure 1 may be followed by a minimum gain constraint module 108.

Figure 2 illustrates another system for sound signal processing. One or more input signals 101 are passed to a channel separator 102. The input signals are usually provided by one or more microphones or by signals transmitted from a remote microphone as in a telephone for example or from a signal store as in an MP3 player for example. The channel separator 102 provides means for optionally separating the input signal(s) into parallel signal channels for further processing by one or more channel processors 203. In one embodiment of the present invention, the channel separator 102 comprises a bank of bandpass filters and each channel 203 processes the output of one bandpass filter. In another embodiment of the present invention, the channel separator 102 splits the signal into one or more slowly varying parts (the background noise) and one or more rapidly varying parts (the desired signals). Each channel processor 203 processes its input signal to produce an output signal that is passed to the channel combiner 205. The channel combiner 205 sums the output signals of the channel processors 203 to form one or more output signals 107 from the sound processor. The output signals 107 are usually converted to sound or transmitted to a remote location where they are converted to sound by a loudspeaker or similar transducer. The channel processors 203 may optionally receive or send control signals 204 from or to other channel processors 203. Each channel processor 203 in the embodiment illustrated by Figure 2 may comprise an example of the present invention with other signal processing means.

Figure 7 shows an example architecture for a channel processor 103 or 203 including the present invention and an input-based automatic gain control (AGCi). The input signal 701 to the channel processor comes from the channel separator 102. The amplitude estimator 703 estimates the amplitude of the input signal 701 and passes the amplitude estimate to the AGCi module 704 and the output limiter 706. The AGCi module 704 calculates an initial gain value using an input-gain function as shown, for example in the low-to-mid intensity region of Figure 10. The initial gain value is passed to the minimum gain constraint module 705 which constrains the gain value to be greater or equal to the minimum gain value stored in the minimum gain constraint module. The constrained gain value is passed to the output limiter 706 which constrains the product of the amplitude estimate and the gain value to be less than or equal to a maximum output limit stored in the output limiter 706. The output limit constraint is achieved by reducing the gain value 707 if required. In an off-line signal processor embodiment, as illustrated in Figure 1, the gain value 707 is passed from the channel processor 103 to the adaptive processor 106 as one of the control signals 105. In an in-line signal processor embodiment, as illustrated in Figure 2, the gain value 707 multiplies the output of the optional unity gain processing module 702 using multiplier 708 and the result of the multiplication is the output 709 of the channel processor 203 which is passed to the channel combiner 201.

Figure 8 shows an example architecture for a channel processor 103 or 203 including the present invention and an output-based automatic gain control (AGCo). The input signal 701 to the channel processor comes from the channel separator 102. The input signal 701 is optionally processed by the unity gain processing unit 702 and the output of the unity processing unit is multiplied by the gain value 707. The output amplitude estimator 803 estimates the amplitude of the output signal 709 and passes the amplitude estimate to the AGCo module 804 and the output limiter 706. The AGCo module 804 calculates an initial gain value using an output-gain function as shown, for example in the low-to-mid intensity region of Figure 1 1. The initial gain value is passed to the minimum gain constraint module 705 which constrains the gain value to be greater or equal to the minimum gain value stored in the minimum gain constraint module. The constrained gain value is passed to the output limiter 706. The output limiter 706 compares the amplitude estimate from the amplitude estimator 803 with the maximum output level stored in the output limiter. If the output amplitude estimate is greater than the output limit, the gain value 707 is reduced. If the output amplitude estimate is less than the output limit, then the gain value 707 is replaced by the lesser of the constrained gain value from the minimum gain constraint module 705 and the gain value 707 multiplied by the ratio of the maximum output limit and the output amplitude estimate. In an off-line signal processor embodiment, as illustrated in Figure 1, the gain value 707 is passed to the adaptive processor 106 as one of the control signals 105. In an in-line signal processor embodiment, as illustrated in Figure 2, the gain value 707 multiplies the output of the optional unity gain processing module 702 using multiplier 708 and the result of the multiplication is the output 709 of the channel processor 203 which is passed to the channel combiner 201.

The input-output function for the channel processors 103 and 203 described above will resemble the input-output function illustrated by Figure 9. The advantages of these embodiments of the present invention include the similarity of the input-output function of the channel processor with the normal input-output function for normal hearing illustrated by Figure 6; the linearity of the input- output function in the mid-to-high intensity range which provides greater naturalness and sound quality; the capacity to set a suitable value for the minimum gain without requiring a change in the AGCi, AGCo or maximum output limit settings; the capacity to use high compression ratios and faster time constants than would otherwise be possible in the low-to-mid intensity region where the distortion artefacts generated by the compression will be less audible than if compression was applied in the mid-to-high intensity range; and the stapedius reflex for loud sounds and the vocalization- induced stapedius reflex will have their normal effects.

It is a further advantage of the invention that conventional fitting prescriptions for hearing aids can easily be modified for the fitting of devices that include the present invention. Non-linear fitting prescriptions usually prescribe a gain for soft sounds and a gain for loud sounds, or parameters such as kneepoints and compression ratios that allow calculation of gains for arbitrary input levels. The minimum gain value for the invention should be approximately equal to the gain prescribed for loud sounds, and the low-to-mid intensity parts of the input-output function should be as normally prescribed. The compression region should be just above the hearing threshold, and the compression ratio should be higher than is normally prescribed to allow for a reasonably large linear region in the mid-to-high intensity part of the input-output function. The limiting region of the input-output function should be as normally prescribed.

It is a further advantage of the invention that the low-to-mid intensity part, the mid-to-high intensity part, and the limiting region of the input-output function are completely specified by parameters that are independent of one another and can be used in fine tuning the hearing aid to correct problems for soft sounds, moderately loud sounds, and very loud sounds respectively without causing new problems in other regions.

In the case of a consumer audio device, the compression region should be placed just above the ambient noise level in order to make soft sounds more audible against the background noise. The minimum gain value should be sufficiently large to achieve an acceptable signal-to-noise ratio for the average level of the signal. A manual volume control or an automatic volume control may be suitable for adapting the minimum gain value for the present invention. Alternative methods may be used to adapt the parameters of the input-output function, one example of which being the technique set out in PCT/AU2004/001691, the content of which is incorporated herein by reference.

A further embodiment of the present invention comprises an ADRO channel processor as set out in US Patent No. 6,731,767, the content of which is incorporated herein by reference. Figure 12 illustrates and ADRO channel processor comprising and ADRO processor according to PCT/AU2004/001691 and a minimum gain constraint module 1207 according to the present invention. The ADRO channel processor illustrated by Figure 12 can be used in either an off-line or in-line architecture as illustrated by Figures 1 and 2 respectively. The input signal 1201 to the ADRO channel processor comes from the channel separator 102. The input signal 1201 is multiplied by the gain value 1210 by multiplier 1202. The amplitude estimator 1203 estimates the amplitude of the output signal 1211 and passes the amplitude estimate to the percentile estimator 1204 and the output limiter 1209. The percentile estimator 1204 calculates a high percentile value, for example the 90^th percentile which is the amplitude value that is exceeded 10% of the time, and a low percentile, for example the 30^th percentile which is the amplitude value that is exceeded 70% of the time. The two percentile values are passed to the gain adjuster 1205. If the high percentile is higher than the "comfort target value" stored in the gain adjuster 1205, the gain adjuster decrements the interim gain value 1208, otherwise if the low percentile is below the "audibility target vale" stored in the gain adjuster, the gain adjuster increments the interim gain value by a small amount. These steps are called the "comfort rule" and the "audibility rule" of ADRO. The interim gain value is compared with a maximum gain value stored in the maximum gain constraint module 1206. If the interim gain is greater than the maximum gain value, the interim gain is set equal to the maximum gain value. This step is called the "background noise rule" of ADRO. The interim gain value is compared with a minimum gain value stored in the minimum gain constraint module 1207. If the interim gain is less than the minimum gain value, the interim gain is set equal to the minimum gain value. The interim gain value is passed to the output limiter 1209. The output limiter 1209 compares the amplitude estimate from the amplitude estimator 1203 with the maximum output level stored in the output limiter. If the output amplitude estimate is greater than the output limit, the gain value 1210 is reduced by a factor equal to the maximum output level divided by the amplitude estimate. If the output amplitude estimate is less than the output limit, then the gain value 1210 is replaced by the lesser of the interim gain value 1208 and the gain value 707 multiplied by the ratio of the maximum output limit and the output amplitude estimate. This step is called the "hearing protection rule" of ADRO. In an off-line signal processor embodiment, as illustrated in Figure 1, the gain value 1210 is passed to the adaptive processor 106 as one of the control signals 105. In an in-line signal processor embodiment, as illustrated in Figure 2, the output signal 1211 of the channel processor 203 is passed to the channel combiner 201.

In the case of an ADRO sound processor, the present invention introduces a new rule. ADRO is an adaptive linear system with a very slow adaptation rate. This means that ADRO already encompasses most of the advantages that stem from linearity in the mid-to-high intensity range. One advantage of introducing the new ADRO rule is that the minimum gain can be used to prevent ADRO from becoming too soft in the presence of loud noise with a narrow intensity range. This problem can arise for some ADRO hearing aid users under some circumstances. Another advantage is that the minimum gain and the maximum gain concepts are already familiar to audiologists who fit compression hearing aids, as the gain for soft sounds and the gain for moderately loud sounds. By using these parameters, conventional compression fitting prescriptions can also be used for ADRO.

For both compression and ADRO embodiments of the present invention, the optimal minimum gain value may be larger than 0 dB. Notable examples include hearing aid fittings for people with a conductive or mixed hearing loss. In this case, the minimum gain should be set equal to or less than the air-bone gap in the listener's audiogram.

Some portions of this detailed description are presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the means used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is here, and generally, conceived to be a self-consistent series of steps leading to a desired result. The steps are those requiring physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.

As such, it will be understood that such acts and operations, which are at times referred to as being computer-executed, include the manipulation by the processing unit of the computer of electrical signals representing data in a structured form. This manipulation transforms the data or maintains it at locations in the memory system of the computer, which reconfigures or otherwise alters the operation of the computer in a manner well understood by those skilled in the art. The data structures where data are maintained are physical locations of the memory that have particular properties defined by the format of the data. However, while the invention is described in tyhe foregoing context, it is not meant to be limiting as those of skill in the art will appreciate that various of the acts and operations described may also be implemented in hardware.

It should be borne in mind, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise as apparent from the description, it is appreciated that throughout the description, discussions utilizing terms such as "processing" or "computing" or "calculating" or "determining" or "displaying" or the like, refer to the action and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices.

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the scope of the invention as broadly described. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

Claims

CLAIMSThe invention claimed is:

1. An apparatus for processing a sound signal including: input means for receiving electrical or digital representations of sound signals; channel separation means for dividing each represented sound signal into at least one signal channel; means for determining a gain value for each signal channel; means for constraining the gain value for each signal channel to remain at or above a minimum gain value for mid-to-high intensity signals; means for setting, storing, and adjusting the minimum gain value for each signal channel; means for multiplying the input level by the gain value to obtain the output level in each signal channel; means for constraining the output level of each signal channel to be no greater than a predetermined maximum output level; means for combining the outputs of the signal channels; and means for converting the combined electrical or digital representation of the output sound signal to an acoustic sound signal.

2. The apparatus according to claim 1, wherein the means for dividing each represented sound signal into at least one signal channel, the means for dividing each represented sound signal into at least one signal channel, the means for determining a gain value for each signal channel, the means for constraining the gain value for each signal channel to remain at or above a minimum gain value for mid-to-high intensity signals, the means for setting, storing, and adjusting the minimum gain value for each signal channel, the means for multiplying the input level by the gain value to obtain the output level in each channel, the means for constraining the output level of each channel to be no greater than a predetermined maximum output level, and the means for combining the outputs of the signal channels, are implemented by a programmed microprocessor coupled to memory storage means.

3. The apparatus according to claim 2 wherein the means for dividing each represented sound signal into at least one signal channel is based on a frequency analysis that divides the signal into at least one frequency band for each signal channel.

4. The apparatus of claim 3 wherein the means for combining the outputs of the signal channels is to construct a finite impulse response filter with filter taps that provide the required gain value in each frequency band and apply the finite impulse filter to the undivided input signal.

5. The apparatus according to claim 3 wherein the means for determining the gain value for each signal channel is based on an input compression function relating the gain value to the amplitude of the signal at the input to the signal channel.

6. The apparatus according to claim 3 wherein the means for determining the gain value for each signal channel is based on an output compression function relating the gain value to the amplitude of the signal at the output to the signal channel.

7. The apparatus according to claim 4 wherein: the microprocessor is programmed to calculate and store in memory, distribution values indicative of the distribution of the output levels of the plurality of signal channels over a period of time; the means for determining the gain value for each signal channel is based on one or more distribution values which are approximately the 10^th, 30^th, 70 , 90^th and 98^th percentiles of the output level distributions of the plurality of signal channels over a period of time; and the gain values are constrained to be no greater than a predetermined maximum gain value in each signal channel.

8. The apparatus according to claim 7 wherein the minimum gain value is set approximately equal to 12, 15, 18, 21, or 24 dB lower than the maximum gain value in each signal channel.

9. The apparatus according to claim 2 wherein: the means for dividing each represented sound signal into at least one signal channel is based on a frequency analysis and a further analysis to separate frequency component signals coming from desired sound sources, such as speech, from background noise sources such that there are desired signal channels and noise signal channels.

10. The apparatus according to claim 9 wherein: the microprocessor is programmed to calculate and store in memory, distribution values indicative of the distribution of the output levels of the plurality of speech signal channels over a period of time; the means for determining the gain value for each speech signal channel is based on one or more distribution values which are approximately the 30^th, 70^th, 90^th and 98^th percentiles of the output level distributions of the plurality of desired signal channels over a period of time; and the gain values are constrained to be no greater than a predetermined maximum gain value in each signal channel.

1 1. The apparatus according to claim 10 wherein: the microprocessor is programmed to set the gain value for each noise signal channel no higher than the minimum gain value for the desired signal channel at the corresponding frequency in order to increase the signal-to-noise ratio in every frequency band in the combined output signal.

12. The apparatus according to claim 9 wherein the analysis to separate component signals coming from different sound sources assigns rapidly varying components to the desired signal channels and slowly varying components to the noise signal channels.

13. The apparatus according to claim 9 wherein the analysis to separate component signals coming from different sound sources uses independent component analysis and blind source separation techniques in the prior art followed by sound categorization in the prior art to separate desired signal channels from noise signal channels.

14. The apparatus according to claim 9 wherein the analysis to separate component signals coming from different sound sources uses directional information from one or more directional microphones or an array of directional microphones to assign sounds from the front direction to the desired signal channels and sounds from the rear and side directions to the noise signal channels.