FR2610162A1 - Improved auditory prosthesis and method including application thereof - Google Patents

Abstract

This prosthesis comprises a microphone 12, a preamplification stage 18, filtering means 1 for dividing an input signal into several predetermined frequency bands, means 1 of selective compression of the dynamic range of the input signal filtered in each frequency band, means 1 of attenuation/amplification of the signal filtered and compressed in each frequency band, an amplification stage 27 and a transducer 13. The means for filtering, for compression and for attenuation/amplification comprise a digital signal processor 1 connected to the preamplification stage by way of an analog/digital converter and to the amplification stage by way of a digital/analog converter, and at least one programmable memory 10 associated with the signal processor and in which are stored parameters representative of the filtering, compression and attenuation/amplification characteristics specific to the said prosthesis. Application to the hearing-impaired afflicted with partial deafness.

Description

 The invention relates to a hearing aid and a method for correcting the audiogram of a subject, comprising an application.

 The acoustic dynamics of a subject is determined by the interval between the threshold of perception of sounds and the threshold of pain. These thresholds are expressed in acoustic decibeic, the zero decibel value corresponding to the lower limit of sound perception in an individual with good hearing (perception threshold). In contrast, the painful threshold of the ear corresponds to the upper limit of bearable hearing. In subjects with normal hearing, the threshold of perception and the threshold of pain are substantially the same regardless of the frequency of the sound stimulus between 200 and 8000 Hertz. In the hearing impaired, a large number of hearing impairments are due to more or less significant changes in the thresholds of perception and pain depending on the frequency. This is, for example, that a subject presenting at 1000 Hertz a threshold of perception of 40 dbA and a pain threshold of 60 dBA will see at this frequency its dynamic reduced 20 dBA while a normal value is around 80 dBA.

 Multichannel hearing aids try to remedy these deficiencies by ensuring processing of the received signal which is different according to the frequencies.

Such a prosthesis includes a microphone converting a sound signal into an electrical input signal, a stage for pre-amplifying the input signal, and filtering means for dividing the input signal into several predetermined frequency bands. In known multi-channel prostheses, and each filter determining a frequency band is associated with a processing channel comprising means for selective compression of the dynamics of the filtered input signal and means for attenuation or amplification of the filtered and compressed signal . The signals from the different channels are then mixed to be amplified by a power amplifier producing an output signal driving a transducer which restores a processed sound signal.

 Multichannel hearing aids of this type are described in particular in documents US-A-4,025,723 and DE-A-2641675. Patent FR-A-2566658 describes an improvement to these multi-channel prostheses which consists in further providing a preliminary compression stage situated downstream of the pre-amplification stage and upstream of the filtering means and to be given to the filter of each tracks a slope of predetermined characteristics.

 These multi-path prostheses of the state of the art and the methods that they implement nevertheless have the drawback of comprising a limited number of treatment paths and, above all, of offering filtering, compression and attenuation characteristics. / amplification in each channel which cannot be adapted to the specific needs of each patient except to present prohibitive costs.

 The present invention aims to produce an improved hearing aid which makes it possible to easily adapt the filtering, compression and attenuation / amplification characteristics to each patient.

To this end, the invention relates to a hearing aid comprising
- at least one channel for acquiring and converting a sound signal into an electrical input signal,
filtering means for dividing the input signal into several predetermined frequency bands,
means for selective compression of the dynamic range of the filtered input signal in each of said frequency bands,
means for attenuation / amplification of the filtered and compressed signal in each of the frequency bands, and
- at least one emission channel producing a sound signal processed from the signal from the attenuation / amplification means,
characterized in that the filtering, compression and attenuation / amplification means comprise
a digital signal processing processor connected to the acquisition channel via an analog / digital converter and to the transmission channel via a digital / analog converter, and
at least one programmable memory associated with the signal processing precursor and capable of storing parameters representative of the filtering, compression and attenuation / amplification characteristics specific to said prosthesis.

 Preferably, the prosthesis comprises means of connection to an external tool for programming the parameters in said memory. An audioprosthetist can thus, by means of a programming tool such as a microcomputer, directly program a prosthesis to adapt it to the specific characteristics of the audiogram which he will have detected on a patient. The use of a programmable and electrically erasable memory will also make it possible to modify at any time the parameters previously recorded to take account of the modifications which would have occurred in the patient's audiogram. In this regard, the use of a NOV memory RAM presents more particularly the advantage of combining a read only memory area and a random access memory area and of being addressable by a serial link from the programming tool, thus allowing the use of simple and inexpensive connection means. .

The invention also relates to a method for correcting the audiogram of a subject by means of a prosthesis as defined above, characterized in that
- the input signal is sampled and converted into digital form,
- the Fourier transform of said sample is calculated on a predefined frequency spectrum,
- adapting said transformed sample to the subject's audiogram by ensuring & each frequency identified in said spectrum a frequency transposition in the same spectrum by means of two square matrices of coefficients C <ij and N ij determined experimentally for said subject and representative respectively its perception threshold and its pain threshold each of two frequencies of said spectrum,
said sample thus treated is converted into analog form, and
- converting said analog signal into an audible signal.

Other characteristics and advantages of the invention will emerge from the description which follows of an embodiment given solely by way of example and illustrated by the appended drawings in which
Figure 1 is a block diagram of a digital hearing aid according to the invention
Figure 2 is a block diagram on which the acquisition and emission stages of Figure 1 have been shown in more detail
FIG. 3 is a timing diagram illustrating the operation of the diagram in FIG. 2.

FIG. 4 is a table giving an example of the distribution of the main frequencies used for processing a sound signal;
FIG. 5 is a diagram illustrating an example of correction of the audiogram of a subject; and
Figure 6 is a block diagram similar to that of Figure 2 and illustrating an alternative embodiment.

 The prosthesis of FIG. 1 comprises a digital signal processing processor 1 which can be, for example, the TMS320 circuit from the company Texas Instruments.

 A data bus 2 connects the processor 1 to a transmission stage 3, to an acquisition stage 4 via a buffer circuit 5, to a second buffer circuit 6 and a locking circuit 7. The processor 1 is also connected to circuits 5, 6 and 7 by an address bus 8 in which a decoding circuit 9 is interposed.

 The prosthesis itself is completed by a programmable memory NOV RAM 10 having a serial input on a connector 11 and a serial output connected to the buffer circuit 6. In memory 10 can be stored parameters which determine the characteristics of the treatment carried out by processor 1 as a function of the program contained in a read-only memory which, in the representation of FIG. 1, is assumed to be part of the processor. Connector 11 terminals are also connected to memory control inputs 10 and to the locking circuit 7.

 The acquisition stage 4 is connected to a microphone 12 and the emission stage 3 to a transducer 13 such as an earphone.

 A volume control member 14 and a function selection member 15 intended to be handled by the patient are connected to the buffer circuit 6. Finally, display means 16 such that one or more light-emitting diodes are driven by an output of the locking circuit 7.

 Before any use, the prosthesis of figure 1 must be programmed according to the hearing characteristics of the patient who will be equipped with it. A doctor can easily carry out this operation himself by means of a programming tool such as a microcomputer 17, connected to the connector 11, which will load into the memory NOV RAM 10 the parameters determining the filtering characteristics (slope, bandwidth), dynamic compression and attenuation / amplification in each frequency band, for example, a high number of narrow band filters can be used in each frequency domain where the patient has hearing impairments in order to finely process the signal, while it will be enough to combine several filters or to use a small number of them to carry out broadband filtering in the frequency range where the audiogram of the patient is suitable.

 The filtering is ensured by the processor 1 by digital filtering or fast Fourier transform (FFT) by means of algorithms well known to those skilled in the art which take place as a function of the instructions contained in the program read-only memory parameterized by the programmable memory 10. The parameters contained in the latter also determine the rate of compression of the dynamics as well as the rate of attenuation and / or amplification which will be applied to the signals filtered digitally in each of the frequency bands.

 Once the prosthesis is programmed according to the hearing characteristics of the patient, the operation is as follows: the microphone 12 picks up the sound signals and converts them into an analog input signal which is amplified, filtered and converted into digital form by the acquisition stage 4. The buffer circuit 5 makes it possible to store a sample converted into digital form until it is read by the processor 1.

 The latter then processes this sample by performing the filtering, compression and attenuation or amplification operations.

First, the digitized signal is transformed by FFT in order to calculate the projection of the sampling (S) on a predefined frequency spectrum fl .... fn. fl to fn are the main frequencies chosen during the design of the prosthesis. We have
S = g ai fi, ot ai is the coefficient assigned to each frequency fi of the Fourrier transform.

To each identified frequency fi are applied correction coefficients precharacterized during programming: these correction coefficients are gathered in a correction matrix & two dimensions (n rows and n columns). This multifrequency auditory correction matrix provides for each pure frequency fi identified during the calculation by FFT a specific projection in the same frequency field fl fn. If n is the total number of frequencies fi identifiable by calculation by FFT, this matrix is defined by coefficients &alpha; ij such that, if i is the order of the frequency to be corrected, whatever i is, the projection of aifi is
j = n
ai ZW ijfj
j = 1
In a specific embodiment, n is arbitrarily linked to sixteen. Furthermore, the correction matrix is limited to a diagonal matrix, that is to say that all the coefficients aij such that i # j are zero. Whatever i is, the projection of aifi is therefore i. have. fi.

 Thus, no frequency transposition is carried out in this specific embodiment.

For each frequency, only the hearing threshold is raised by the correction coefficient cRi to ensure adaptation to the sensitivity of the patient's ear.

 The table in FIG. 4 gives an example of the distribution of the main frequencies fi.

 In order to take into account the dynamic possibilities of the ear (modification of the painful perception curve), a precharacterized weighting coefficient / 6ij is defined for each main frequency fi.

These coefficients ss ij are also collected in a correction matrix with n rows and n columns similar to the matrix of coefficients q ij. The two matrices are linked: if there exists j such that CXij = O, then ssij = O.

 In the specific embodiment mentioned above, the matrix of coefficients ss ij is also diagonal.

 There are therefore sixteen coefficients ss i, that is to say that in total thirty two coefficients are programmed by the hearing care professional.

 Whatever the number n of main frequencies chosen, it is obvious that the processing carried out by the processor 1 will be all the faster as the number of the coefficients stAij and ss ij will be low. The total number will be at least 2n if we want to restore any sample on the whole of the predefined spectrum, which corresponds to the case of two diagonal matrices.

 FIG. 5 shows an example of a correction for normalizing the audiogram of a patient. The acoustic level expressed in dBA is represented on the predefined spectrum fl .... fn. Some coefficients O <i and S 'have been shown by way of illustration. The curves in broken lines SPN and SDN represent the thresholds of perception and pain of a subject endowed with normal hearing, the curves in solid lines SPC and SDC the same thresholds adapted to a patient presenting hearing impairments.

 The processed samples are applied to the emission stage 3 which reconstitutes an analog signal applied to the transducer 13. The latter delivers to the patient an audible signal whose dynamics and amplitude in each frequency band are perfectly adapted to his own hearing characteristics.

 In addition, the patient will be able to adjust the overall level of the sound signal emitted by means of the volume control member 14: the selected value applied to the buffer circuit 6 is read thereon by the processor 1 which controls the amplification rate. of the power amplifier integrated into the transmission stage 3.

 Furthermore, the function selection member 15 has an inactive position and two active positions, one of which has the role of ensuring attenuation of background noise in the absence of a sound signal picked up by the microphone. As for the volume control, the processor 1 reads the information corresponding to the output of the buffer circuit 6 and consequently controls the aforementioned power amplifier.

In its second active position, the function selection member triggers a prosthesis operating test which makes it possible to detect one of the following three faults:
- interruption of the electrical connection between the acquisition stage 4 and the microphone on the one hand and between the emission stage 3 and the transducer 13 on the other hand;
- insufficient state of charge of the battery
- malfunction in the electronic circuits of the prosthesis.

 Any defect in one of the three categories above will be signaled to the patient by means of display means 16, for example by a characteristic lighting of one or more light-emitting diodes. This ignition is controlled by the processor 1 via the locking circuit 7.

 The block diagram of FIG. 2 is a partial representation of the prosthesis of FIG. 1, on which the emission 3 and acquisition 4 stages have been more detailed.

 The acquisition stage 4 comprises an amplification and adaptation circuit 18 through which the electrical input signal from the microphone 12 is applied to an analog bandpass filter 19. The role of the filter 19 is to keep from the input signal only the frequency band falling in the range of the audible spectrum. The output signal of the bandpass filter 19 drives the input of a sampler-blocker 20 whose output is connected to the input Vin of an analog / digital converter 21. The circuits 18, 19, 20 and 21 constitute the acquisition stage 4 of FIG. 1.

 The outputs of the analog / digital converter 21 are connected to the inputs of the buffer circuit 5. The outputs of the latter are connected by the data bus 2 to the data inputs D of the processor 1. The buffer circuit 5 can be constituted for example by a component 74HCE 245.

The signal processing processor 1 is associated with a read only memory 22 in which is stored the configurable program which is executed by the processor 1 to perform the filtering, compression and attenuation / amplification functions on the samples received from the buffer circuit 5. The processor 1 and memory 22 are connected to each other, in particular by the data bus 2 and by an address bus 23. The parameters which determine the characteristics of the processing specific to each subject are read by processor 1 in the NOV memory
RAM 10 via buffer circuit 6.

 The data bus 2 is also connected to the inputs of a locking circuit 24 whose outputs drive the inputs of a digital / analog converter 25. The locking circuit can be formed from flip-flops D. The converter 25 converts the processed samples which are applied to it in an analog signal present on its output Vs which attacks an analog bandpass filter 26. This filter 26 is intended to restore an analog signal whose frequency band is located in the audible spectrum. The filtered signal coming from the filter 26 drives a power amplifier 27. The circuits 24, 25, 26 and 27 constitute the emission stage 3 of FIG. 1. The output of the power amplifier 27 is connected to the transducer 13 consisting, for example, of an earpiece.

 In the block diagram of FIG. 1, the buffer circuits 5 and 6, the locking circuit 7 and the transmission stage 3 are addressed by the processor 1 by means of an address bus 8 and a circuit for decoding addresses 9. This bus 8 and the circuit 9 have not been shown in FIG. 2 because, in this particular embodiment, the signal processing processor 1 is assumed to be a TMS circuit 320. The latter has the particularity of having specific inputs-outputs which allow direct control of some of the circuits of the acquisition stage 4 and of the transmission stage 3. Furthermore, the processor 1 is controlled by a base of time 28.

 The output WE of processor 1 is connected 9 the input CP of the locking circuit 24. The output DEN of processor 1 is connected, on the one hand & the input E of the buffer circuit 5 and, on the other hand, & l 'trigger input Io of a monostable 29. The output Q of the monostable 29 is connected to one of the inputs of an OR gate whose output is inverted and connected to input B & of the analog / digital converter 21 A second time base or clock 31 controls the sampler-blocker 2O as well as a second monostable 32. The Q output of this monostable 32 is connected to the Io input of a third monostable 33 whose Q output is connected at the second entrance to OR gate 30.

 The operation of the circuit of FIG. 2 will now be described with reference also to the timing diagram of FIG. 3. When the microphone 12 picks up a sound, the corresponding analog signal is applied to the sampler-blocker 20 after amplification-adaptation and filtering in 18 and 19.

 On the rising edge of the clock signal 31 (FIG. 3A) the blocking sampler 20 begins the acquisition of a new sample. The duration of the sampling will depend on the capacity of the charge capacitor (not shown) of the sampler-blocker 20 and may be between, for example, approximately 4 and 20 microseconds. The monostable 32 introduces a delay (FIG. 3B) relative to the rising edge of the signal from the clock 31 so that the conversion cannot start until the sampling is finished. As soon as the Q output of the monostable 32 goes & O (Figure 3B), the Q output of the monostable 33 goes & 1 (Figure 3C). The reverse output of the OR gate 30 and the input B & of the analog / digital converter 21 pass & O (FIG. 3D) and the conversion begins.

Once the conversion is complete, the analog / digital converter 21 delivers on its output DR (FIG. 3E) a signal which activates the input BIO of the processor 1 (FIG. 3F). This signal causes the execution of an instruction
Interrupt IN which activates the DEN output (figure 3G) of processor 1.

 The processor 1 then reads the converted data present on the bus 2 at the output of the buffer circuit 5 as long as the output DEN is at the low level. The falling edge of DEN also has the role of triggering the monostable 29.

The Q output of the monostable 29 must go to the high level (FIG. 3H) before the Q output of the monostable 33 goes to the low level so that the input B & remains in the low state, that is to say that the the converted signal is maintained at the output of the analog / digital converter 21, as long as the buffer circuit 5 (three-state amplifier) is on.

This ensures that the signal read by processor 1 is the converted signal. As soon as the output Q of the monostable 29 returns to the low level, the signal B & rises to the high level and the converted sample disappears because the output DR rises to the high level just after B &.

 If an AD570 circuit is used for the analog / digital converter 21, the time constant of the monostable 33 must be greater than 25 microseconds.

The time constant of the monostable 29 must be slightly greater than that during which the signal DEN is at the low level, duration representing the time of acquisition by processor 1 of the converted sample. If the time constant of the monostable 29 is equal to 10 microseconds, the sampling period determined by the signal from the clock 31 can be equal to or greater than 40 microseconds.

 When the output Q of the monostable 29 is returned to the low level, the processor 1 begins the digital processing of the converted sample. The processing program ends with a data output request which activates the output Wu of the processor 1. Under this command, the data present at the input of the locking circuit 24 is transmitted and maintained at the input of the digital converter / analog 25. The analog signal thus obtained at the output of converter 25 is filtered at 26 then amplified at 27 to be applied to the earpiece 13.

 A new cycle of sampling, conversion into digital form, acquisition and processing by the processor 1 takes place on the appearance of the next rising edge of the signal from the clock 31 (FIG. 3A).

According to the example of embodiment described, a dioprosthetist can program directly in the prosthesis the parameters specific to a patient according to
According to the ahcriz embodiment, an audioprosthetist can program the parameters specific to a patient directly into the protonese according to the audiogram thereof. In particular, the distribution of the filter bank in the audible spectrum of the patient, ie the compression rate applied in each of the frequency channels or bands determined by each filter as well as the level of attenuation / amplification in each of these same bands. selectively optimized and reconfigured if the hearing care professional notices a change in the patient's audiogram over time.

 FIG. 6 shows an alternative embodiment which makes it possible to carry out a differentiated treatment for each ear of a patient. This variant differs from the example in FIG. 2 by the fact that it has two emission stages instead of one and the same reference numbers have been kept, the emission stages and the circuits which constitute them having been assigned letters G and D depending on whether they relate to the left or right ear.

 The detailed description of the diagram in FIG. 6 and of its operation will therefore not be repeated. It will simply be noted that the processor 1 is assigned two outputs WEG and WED making it possible to selectively transmit the data present on the bus on the 3G stage or on the 3D stage. The processor performs, for example, successively two treatments for each sample read at the output of the buffer circuit 5 and the two processed samples are selectively directed to the 3G and 3D stages as indicated above and, from there, to the right earphones and left 13G and 13D.

 To carry out this differential processing, at least some of the parameters contained in the memory NOV RAM 10 are duplicated and provided with an assignment code. Thus, for example, two matrices of coefficients C <ij and two matrices of coefficients <ij can be stored, and each assigned to one of the two right and left transmission channels. In the case of the specific embodiment described above, it is then 64 coefficients, 32 for each ear, that the hearing care professional can program.

 This variant makes it possible to further improve the performance and flexibility of adaptation of the prosthesis because it is common for subjects to present different audiograms for each ear.

 In order to facilitate the procedure for programming the coefficients cti and n i for a given patient, a procedure may consist in having the hearing aid acoustician correct the hearing by modifying the patient's audiogram.

 Thus, for example, the patient's audiogram being that represented in FIG. 5 by the solid line curves SPC and SDC, the hearing care professional will propose perception threshold and pain threshold curves which may be those represented in broken lines (SPN, SDN).

 To this end, the hearing care professional will use the programming tool 17 equipped with conversational software, a graphic input device and a table to digitize conventionally used as a computer device. This table will allow the digitization of the patient's actual audiogram and its display on the screen of the tool 17. On this screen will also be displayed the SPN and SDN curves of the reference audiogram which either will be defined case by case by the hearing care professional, or will preferably be stored in tool 17 with the scales (frequency, dBA) of the audiogram.

Alternatively, if the audiometric measurement equipment has a computer communication means (terminal
VT 100 for example), the programming tool 17 can be directly connected to it and the use of a digitizing table will not be necessary.

 The hearing care professional will directly define point by point on the screen of the tool 17 the corrections to be made to the patient's audiogram. I1 can carry out these corrections for each of the n frequencies of the predefined frequency spectrum (n = 16 in the particular example of FIG. 4) by moving the point of the curve SPC on the curve SPN and the point of the curve SDC on the curve SDN using the graphic input device (photostyle or "mouse" for example).

The programming tool 17 will then calculate directev; - the coefficients Xi and ss i represented by the amplitude of the displacement of the patient's audiogram at each of the frequencies of the predefined spectrum and, if necessary, will then smooth the curve corrected points in order to allow the hearing aid acoustician to verify that the corrected audiogram substantially coincides with the reference audiogram, before it validates the coefficients o (i and ss i calculated which will be transferred into the memory
NOV RAM 10 of the prosthesis as described previously.

 In the case of a differential treatment for each ear as described in connection with FIG. 6, an audiogram is noted for each ear and the procedure for programming the coefficients described above will be carried out for each audiogram.

 It goes without saying that the embodiment described is only an example and it could be modified, in particular by substitution of technical equivalents, without thereby departing from the scope of the invention. thus, for example, that the programming of the characteristic parameters of the patient could not be carried out by the hearing care professional himself but by the manufacturer of the prosthesis on the basis of the data provided by the hearing care professional. An EPROM memory could then be used in place of the NOV RAM memory.

Claims (16)

 1. Hearing aid including  - at least one channel (12, 4) for acquiring and converting an audible signal into an electrical input signal,  - filtering means (1) for dividing the input signal into several predetermined frequency bands,  - means (1) for selective compression of the dynamics of the filtered input signal in each of said frequency bands,  means (1) for attenuating / amplifying the filtered and compressed signal in each of the frequency bands, and  - at least one emission channel (3, 13) producing a sound signal processed from the signal from the attenuation / amplification means,  characterized in that the filtering, compression and attenuation / amplification means include:  - a digital signal processing processor (1) connected to the acquisition channel (12, 4) via an analog / digital converter and to the transmission channel (3, 13) via the '' via a digital / analog converter (25), and  - At least one programmable memory (10) associated with the signal processing processor and able to store parameters representative of the characteristics of filtering, compression and attenuation / amplification proper to said prosthesis.  2. Prosthesis according to claim 1, characterized in that it comprises means (11) for connection to an external tool (17) for programming the parameters in said memory (10).  3. Prosthesis according to claim 2, characterized in that said memory (10) is programmable and electrically erasable.  4. Prosthesis according to any one of claims 1 & 3, characterized in that said memory (IG) is of the NOV RAM type.  5. Prosthesis according to any one of claims 1 to 4, characterized in that it comprises a program read-only memory (22) determining the characteristics of digital filtering, compression and attenuation / amplification executed by the processor signal processing (1) according to the parameters stored in said programmable memory (10).  6. Prosthesis according to any one of claims 1 to 5, characterized in that it comprises a buffer circuit (5) between the analog / digital converter (21) and the processor (1) and a locking circuit (24) between the processor (1) and the digital / analog converter (25).  7. Prosthesis according to any one of claims 1 to 6, characterized in that the acquisition channel comprises a microphone (12), a pre-amplification stage (18), a bandpass filter (19) and a blocker sampler (20).  8. Prosthesis according to any one of claims 1 to 7, characterized in that the transmission channel (3, 13) comprises said digital / analog converter (25), a bandpass filter (26), a stage d amplification (27), and a transducer (13).  9. Prosthesis according to any one of claims 1 to 8, characterized in that it comprises means (14) for controlling attenuation of background noise in the absence of detected sound signal.  10. Prosthesis according to any one of claims 1 to 9, characterized in that it comprises means (15) for controlling the processor (1) the execution of functional tests of the prosthesis and the means (16) d display of the results of said tests.  11. Prosthesis according to any one of claims 1 to 10, characterized in that it comprises two transducers (13G, 13D) and in that said programmable memory (10) is capable of memorizing parameters allowing the processor (1) perform specific signal processing for each of said transducers.  12. Prosthesis according to claims 8 and 11, characterized in that it comprises two transmission channels (3G, 13G; 3D, 13D) connected to the processor (1) respectively by two locking circuits (24G, 24D) and in that the processor (1) drives the locking circuits (24G, 24D) to selectively transmit the processed data present at its output to one or the other of said transmission channels.  13. Method for correcting the audiogram of a subject by means of a hearing aid according to any one of claims 1 to 12, characterized in that  - the input signal is sampled and converted into digital form,  - the Fourier transform of said sample (S) is calculated on a predefined frequency spectrum (fl fn),  - adapting said transformed sample to the subject's autogram, ensuring at each identified frequency (fi) in said spectrum a frequency transposition in the same spectrum (fl fn) by means of two square matrices of determined coefficients tij and ij experimentally for said subject and representative respectively of their perception threshold and their pain threshold at each of the frequencies of said spectrum,  said sample thus treated is converted into analog form, and  - converting said analog signal into an audible signal.  14. Method according to claim 13, characterized in that said matrices are diagonal such that sc.rtC that, at each frequency identified in said spectrum of the transformed sample, the frequency transposition makes the same frequency correspond.  15. Method according to claims 13 or 14 when they depend on claims 11 or 12, characterized in that two matrices of said coefficients CN i, and ss ij are assigned to each of said transmission channels (3G, 13G; 3D, 13D ).  16. Method according to any one of claims 13 to 15, characterized in that said coefficients CXij and ss ij are determined experimentally using a programming tool (17) on which the patient's audiogram is displayed. and a reference audiogram and the patient's audiogram is moved point by point to make it substantially coincide with the reference audiogram by means of a graphic input device, the programming tool calculating said coefficients &alpha; ij and ss ij represented by the amplitude of the displacement of the patient's audiogram & each of the frequencies of said predefined spectrum.