Patent Application
20240188853
The present invention relates to a method for estimating an audiogram for a specific user. The invention also relates to a hearing estimation system.
Generally, a hearing aid system according to the invention is understood as meaning any system which provides an output signal that can be perceived as an auditory signal by a user or contributes to providing such an output signal, and which has means adapted to compensate for an individual hearing loss of the user or contribute to compensating for the hearing loss of the user. These systems may comprise hearing aids that can be worn on the body or on the head, in particular on or in the ear, or that can be fully or partially implanted. However, a device whose main aim is not to compensate for a hearing loss, for example a consumer electronic device (mobile phones, MP3 players, so-called “hearables” etc.), may also be considered a hearing aid system, provided it has measures for compensating for an individual hearing loss.
Within the present context, a hearing aid can be understood as a small, battery-powered, microelectronic device designed to be worn behind or in the human ear by a hearing-impaired user. Prior to use, the hearing aid is adjusted by a hearing aid fitter according to a prescription. The prescription is based on a hearing test, resulting in a so-called audiogram, of the performance of the hearing-impaired user's unaided hearing. The prescription is developed to reach a setting where the hearing aid will alleviate a hearing loss by amplifying sound at frequencies in those parts of the audible frequency range where the user suffers a hearing deficit. A hearing aid comprises one or more microphones (or, more generally, electroacoustic input transducers), a battery, a microelectronic circuit comprising a signal processor, and an acoustic output transducer. The signal processor is preferably a digital signal processor. The hearing aid is enclosed in a casing suitable for fitting behind or in a human ear.
According to variations, the hearing aid need not comprise a traditional loudspeaker as output transducer. Examples of hearing aid systems that do not comprise a traditional loudspeaker are cochlear implants, implantable middle ear hearing devices (IMEHD), bone-anchored hearing aids (BAHA) and various other electro-mechanical transducer-based solutions including, e.g., systems based on using a laser diode for directly inducing vibration of the eardrum.
Within the present context a hearing aid system may comprise a single hearing aid (a so-called monaural hearing aid system) or comprise two hearing aids, one for each ear of the hearing aid user (a so-called binaural hearing aid system).
In a traditional hearing aid fitting, the hearing aid user goes to a site of a hearing aid fitter (e.g., an acoustician), and the user's hearing aids are adjusted using the fitting equipment that the hearing aid fitter has in his office. Typically, the fitting equipment comprises a computer capable of executing the relevant hearing aid programming software and a programming device adapted to provide a link between the computer and the hearing aid.
The hearing loss of a hearing-impaired person is generally frequency-dependent and may not be the same for both ears. This means that the hearing loss of the person varies depending on the frequency. Therefore, when compensating for hearing losses, it can be advantageous to utilize frequency-dependent amplification. Hearing aids therefore often provide band split filters in order to split an input sound signal received by an input transducer of the hearing aid, into various frequency intervals, also called frequency bands, which are independently processed. In this way it is possible to adjust the input sound signal of each frequency band individually to account for the hearing loss in respective frequency bands.
The frequency dependent adjustment is normally done by implementing a band split filter and a compressor for each of the frequency bands, hereby forming so-called band split compressors, which may be combined to form a multi-band compressor. In this way, it is possible to adjust the gain individually in each frequency band depending on the hearing loss as well as on the input level of the input sound signal in a respective frequency band. For example, a band split compressor may provide a higher gain for a soft sound than for a loud sound in each frequency band.
Traditionally a hearing aid system is fitted—initially or as part of a subsequent fine tuning—based primarily or exclusively on a recorded audiogram for the individual haring aid system user.
Perhaps the most widespread method is based on pure tone tests, where the individual to be tested (i.e. the test person) is presented for a tone at a specific frequency and at first at a very low loudness that most probably is not audible for the test person, where after the loudness is progressively increased until the test person indicates that the tone is audible whereby the hearing threshold may be established, and from that the hearing loss at that specific frequency as compared to normal hearing subjects may be derived. In order to fully characterize the hearing loss, the test may be repeated for other frequencies in the audible range.
This type of test has been offered as online test for many years. Hereby, an individual who suspects possibly having a hearing loss can take the test at home and record their audiogram without having to make an appointment and travel to a hearing care professional. However, this type of test may be time consuming and some users consider the test uncomfortable and annoying, which additionally may lead to a recorded audiogram of low accuracy.
It is therefore an object of the present invention to provide an improved method of quickly estimating an audiogram, which preferably shall be comfortable for the test person.
According to the invention, this object is solved by a method for estimating an audiogram for a specific user, the method comprising the steps of: providing a first plurality of first audiograms from a corresponding first plurality of preferably hearing impaired persons; providing, for each of the first audiograms, an initial value of a weighting factor; selecting a first frequency, out of a first set of frequencies; measuring a hearing threshold of said specific user at said first frequency; performing a first update on each of said weighting factors corresponding to said first plurality of audiograms in dependence on the difference between the value of the respective first audiogram, at said first frequency, and said measured hearing threshold of said specific user at said first frequency; and determining an estimated audiogram for said specific user as a weighted mean of said first plurality of first audiograms based on said weighting factors, according to their respective first update. Embodiments of particular advantage and inventiveness on their own are explained in the dependent claims and in the following description.
The first audiograms may be given by audiograms measured by an audiometer, or may also be given by in-situ audiograms, i.e., by audiograms measured in the presence of a hearing aid in or at the respective ear during measurement. In the following, both kinds of audiograms may be comprised, except as otherwise stated by explicit specification. For each person out of the first plurality, a corresponding first audiogram is provided, hence the corresponding first plurality of first audiograms. The persons of the first plurality, preferably are given by hearing impaired persons; however, also persons without any noticeable hearing impairment may be included, in particular for statistical reasons.
Preferably, each of the first audiograms is representative of the corresponding person's hearing threshold at several frequencies, wherein the first audiograms may also contain interpolations between the measured hearing thresholds. Most preferably, the frequencies at which a respective hearing threshold is measured for generating the audiogram are equal for all persons of the first plurality.
To each of the first audiograms, a weighting factor is associated, and a respective initial value for each weighting factor is provided, i.e., a number of weighting factors corresponding to the first plurality is initialized by said initial values. In a preferred embodiment, the initial values are equal for all weighting factors associated to the different first audiograms. However, different initial values for different first audiograms are also possible, e.g., to include further statistical information.
Preferably, the first set of frequencies corresponds to the measurement frequencies at which the hearing threshold is determined for obtaining the first audiograms. The first frequency may be selected as a fixed frequency, e.g., as a frequency of 1 kHz or as close as possible to 1 kHz within the first set of frequencies, or may also be selected dynamically, by assigning some spread measure to the first audiograms, and evaluating the spread measure in a way such that a possible deviation from the first audiograms may be reduced the most by a corresponding dynamical determination of the first frequency, a corresponding hearing threshold measurement at said dynamically determined first frequency and a respective update of the weighting factors. The first frequency may furthermore be determined based on further information, such as statistical and/or demographic and/or audiological information about the persons corresponding to the first audiograms.
Determining the specific user's estimated audiogram as a weighted mean based on the weighting factors may in particular comprise using directly the respective values of the weighting factors given by their first update. However, and in particular in case of further iterations, the information contained in the first update of the weighting factors is also used for further updates of the weighting factors, thus the notion of a weighted mean based on the weighting factors according to the first update shall also include this case of further updates.
In an advantageous way, the method extracts and uses statistical information contained in the first audiograms, in order to reduce and possibly minimize the number of frequencies at which a true hearing threshold measurement of the specific user needs to be performed in order to obtain a reliable audiogram of said specific user. The first frequency may be selected in a way such that a single hearing threshold measurement at the selected frequency in order to reduce an “a priory uncertainty” about the underlying true audiogram, and thus, the full range hearing ability of the specific user, is performed where said measurement may yield the most relevant information gain with respect to the true audiogram.
In an embodiment, the method further comprises a step of determining, at each of said first set of frequencies, a weighted variance metric for said first plurality of first audiograms, in dependence of said weighting factors, wherein the first frequency is selected out of said first set of frequencies by a determination in dependence on the frequency out of a first frequency range for which the weighted variance metric is largest.
The notion of a weighted variance metric for the first plurality of audiograms, at a given frequency out of the first set of frequencies, shall be understood as a metric for the dispersion or the spread of the different first audiograms' values at the given frequency of interest, satisfying the conditions of a variance. In particular, the weighted variance metric is given by the weighted sum of squared mean deviations of all the first audiograms at the given frequencies, the weighted sum being performed according to the respective initial values of the weighting factors.
The determination of the first frequency, out of said first set of frequencies, in dependence on the frequency out of a first frequency range for which the weighted variance metric is largest, in particular may comprise the case that the first frequency does not equal said frequency for which the weighted variance metric is largest in the first frequency range, e.g., in case of interpolations and/or continuous fittings of the first audiograms and/or of the weighted variance metric over the first set of frequencies. In a preferred embodiment, the first frequency is chosen as the frequency for which the weighted variance metric is largest.
The first frequency range may be chosen such that the first frequency is chosen out of a subset of the first set of frequencies, said subset covering the first frequency range, in order to ensure that the first frequency is chosen in a relevant frequency range from an audiological point of view, e.g., from 500 Hz to 3 kHz, preferably from 1 kHz to 2.5 kHz, or similar. Then, even if the weighted variance metric of the first audiograms may be larger for some frequency out of the first set of frequencies, the first frequency is chosen as the frequency with the maximal weighted variance metric over the first frequency range. However, this condition may also be relaxed, and the first frequency range may also cover the full frequency range, such that the first frequency is chosen in dependence on the frequency of the overall maximum of the weighted variance metric (over all frequencies).
Since the first frequency is determined based on the spread of the first audiograms, it may be selected in a way such that this spread, i.e., the “a priory uncertainty” about the underlying true audiogram, and thus, the full range hearing ability of the specific user, is reduced the most at the selected frequency by a single hearing threshold measurement.
In a preferred embodiment, the method comprises an iteration of the following steps: determining, at several frequencies out of said first set of frequencies, a weighted variance metric for said first plurality of first audiograms in dependence on said weighting factors, according to their most recent update, respectively; determining a distinguished frequency out of said several frequencies, also in dependence on and preferably as the frequency out of a second frequency range for which said weighted variance metric is largest; measuring a hearing threshold of said specific user at said distinguished frequency; performing an update on each of said weighting factors corresponding to said first plurality of first audiograms in dependence on the difference between the value of the respective first audiogram and said measured hearing threshold of said specific user at said distinguished frequency, respectively; and, after finishing said iteration, further comprising the step of determining the estimated audiogram for said specific user as a weighted mean of said first plurality of first audiograms, using said weighting factors according to their most recent update, respectively, for said weighted mean.
The second frequency range may be given by the first frequency range. In particular, the second frequency range may be given by the full frequency range. In case that the first frequency range is not given by the full frequency range, this in particular means that the first frequency is determined from the maximum frequency of the weighted variance metric over a limited frequency range (the first frequency range), and the first update is performed. Afterwards, the distinguished frequency is determined from the overall maximum frequency of the (updated) weighted variance metric.
Preferably, in each iteration run, the weighted variance metric is determined at all frequencies out of said first set of frequencies for which no hearing threshold of said specific user has been measured yet. In particular, this may include that the number of frequencies at which the weighted variance metric has to be determined for the first audiograms, may reduce with each iteration run. The determination of the distinguished frequency, in particular, may be performed in an analogous way to the determination of the first frequency. Thus, conditions and possible embodiments may be transferred from said determination of the first frequency, mutatis mutandis.
Likewise, the update of the weighting factors corresponding to the first audiograms in dependence on the difference between the value of the respective first audiogram and said measured hearing threshold of said specific user at said distinguished frequency, may be performed in an analogous way to the update in dependence on the difference between the value of the respective audiogram and the specific user's measured hearing threshold at the first frequency. Preferably, for each weighting factor, the update may also depend on each of the differences between the respective values of the first audiograms and the hearing thresholds of the specific user already measured before, at their respective frequencies, i.e., at the first frequency, and possibly at distinguished frequencies during earlier iteration runs.
Iterating the procedure increases the accuracy of the audiogram estimation while still allowing for a reduction of hearing threshold measurements. Typically, an audiogram contains hearing threshold measurement at about 8 to 10 frequencies, so an iteration of two or three total runs (including the initial run corresponding to the first frequency) provides a good tradeoff between an optimal accuracy of the audiogram and a quick determination.
In an embodiment, the iteration is finished after completion of a given number of iteration runs or when the weighted variance metric for said first plurality of first audiograms or a variance for said first plurality of first audiograms falls below a given first threshold. Finishing after a given number of total iteration runs—e.g., after the second or third update to the weighting factors and the respective estimation of the specific user's audiogram—ensures that the estimation has a controllable duration and complexity. Finishing the estimation based on a threshold on the weighted variance metric—according to the most recent update to the weighting factors—or on the total variance of the first audiograms allows for ensuring a lower bound on the estimation accuracy.
Preferably, after each iteration run, the estimated audiogram for said specific user is determined as a weighted mean of said first plurality of first audiograms, using said weighting factors according to their most recent update, respectively, for said weighted mean, the estimated audiogram is visualized, and said visualized estimated audiogram presented to a hearing care professional, such as an acoustician or an audiologist, for decision about stopping the iteration. The visualization is preferably performed by means of an appropriate apparatus comprising a screen or a similar displaying device. E.g., a computerized device on which all processing steps of the method are implemented may either comprise a screen or a similar displaying device for visualization, or may be operationally connected to a screen or a similar displaying device. The computerized device may also comprise—and/or be operationally connected to—suitable input means, such as a keyboard or a touch screen, for the hearing care professional to enter a decision to stop the iteration. It can be of advantage for a hearing care professional who is performing the fitting to decide on when the audiogram estimation is sufficiently accurate.
In an embodiment, for each weighting factor, the initial value is chosen in dependence on a demographic similarity of the specific user with the person corresponding to the first audiogram for which the weighting factor is being provided. The demographic similarity may be with respect to a gender and/or an age group and/or an ethnic group and/or a socioeconomic group (e.g., “working class”, “white collar” etc.) and/or a residential settlement type (e.g., “urban”, “rural” etc.). It can be of advantage to assign a stronger weight to first audiograms from persons with a similar demographical background as the specific user, such as a similar age or a similar working background (and thus, an assumed similarity in previous noise exposure).
In an embodiment, the first plurality of first audiograms is chosen as a subset out of a multiplicity of first audiograms from a corresponding multiplicity of preferably hearing impaired persons in dependence on a demographic similarity of the specific person with a person from said multiplicity corresponding to a respective first audiogram. This means in particular that there exists an overall database of first audiograms (the multiplicity), and for the estimation of the specific user's audiogram, the first plurality of audiograms is selected out of the multiplicity by using demographic criteria, in an analogous way as explained above, in order to have first audiograms from persons who are sufficiently “similar” to the specific user with respect to their demographics. This way, it can be assumed that the spread of first audiograms is reduced, and the accuracy of the estimation may be increased even for performing the method by a single run or few iteration runs.
In an embodiment, a starting sound pressure level (SPL) for measuring the hearing threshold of said specific user at a certain frequency is chosen in dependence on the estimated audiogram and/or weighted variance metric at said certain frequency, using the weighting factors according to their most recent update, respectively. Said certain frequency, in particular, may be given the first frequency or by a distinguished frequency during a subsequent iteration run. Typically, in order to measure a hearing threshold of a test person at a given frequency, a probe tone such as a sine tone is produced with a SPL sufficiently high to ensure that even a hearing impaired test person will normally perceive the probe tone. Then, the SPL is slowly reduced (continuously or step-wise) in order to reach the level at which the test person ceases to hear the probe tone. However, adjusting the SPL to the statistics obtained for a certain frequency from the weighted variance metric allows for a faster measurement of the hearing threshold, as SPL ranges where close to everybody (in particular, persons with a similar hearing impairment as the specific user, due to their stronger weighting) is still hearing, may be skipped.
In an embodiment, a continuous fit is performed on the weighted variance metric for all frequencies in the interval spanned by said first set of frequencies, wherein the first frequency is determined from said continuous fit. In particular, this means: the first set of frequencies, as a discrete set, spans a frequency interval. At all frequencies in the first set, the weighted variance metric for the first audiograms is determined, yielding discrete data points in frequency space. Then, a continuous fit such as a polynomial fit is performed over these discrete data points, yielding a curve of a continuous variance function with a continuous frequency dependence. The first frequency may then be derived from said continuous variance function, e.g., as the frequency where said continuous variance function attains its maximum value.
Preferably, in order to obtain a value of a first audiogram at the first frequency, an interpolation and/or a continuous fit is performed on basis of said first audiogram's values at the frequencies of said first set of frequencies. In particular, the first audiograms may be interpolated (e.g., by a linear or logarithmic interpolation of a first audiogram's values between two frequencies) or continuously fit in a way similar to the weighted variance metric in order to obtain values for each of the first audiograms at frequencies not comprised in the first set of frequencies, and in particular, at which no measured values exist for some or any of the first audiograms. This way, values for the first audiograms may also be obtained in case the first frequency is not comprised in the first set of frequencies.
In an embodiment, the method further comprises the steps of: providing a second plurality of second audiograms from a corresponding second plurality of preferably hearing impaired persons, wherein each person out of the first plurality is also comprised in the second plurality of persons, providing, for each of the second audiograms, a value of a weighting factor; determining, at each of a second set of frequencies, a weighted variance metric for said second plurality of second audiograms, in dependence of said weighting factors, wherein the second set of frequencies is a subset said first set of frequencies; and determining said first frequency and/or said distinguished frequency for the first plurality of first audiograms and/or performing said first update on the weighting factors corresponding to the first audiograms also based on the weighted variance metric for said second plurality of second audiograms.
The second set of frequencies may be a proper subset of said first set of frequencies (i.e., there are frequencies contained in the first set but not in the second set of frequencies) or a so-called improper subset (then, the first set and the second set comprise the same frequencies). The first plurality of persons may be equal to the second plurality of persons, or the second plurality of persons may be larger than the first plurality (i.e., some persons of the second plurality are not comprised in the first plurality). The second audiograms may be of the same nature as the first audiograms, or the first audiograms may in particular given by in-situ audiograms, while the second audiograms may be given by “conventional” audiograms.
The method is then partially performed on the second audiograms in the sense that for the second set of frequencies, a weighted variance metric for the second audiograms is provided. The information contained in said weighted variance metric for the second audiograms may then be used to determine the first frequency and/or the distinguished frequency for the first audiograms.
In particular, the first and second pluralities of preferably hearing impaired persons comprise the same persons, wherein for each person out of said first and second pluralities, respectively, the corresponding first and second audiogram are representing the hearing at either of the two ears. This particularly means that the first and second audiograms are measured at the left and right ear (no specific assignment about which audiogram corresponds to which ear shall be implied) of each person out of the first plurality of persons (which are the same persons as the second plurality). Even though hearing loss normally affects both ears of a hearing impaired person in a different way, the ensemble statistics inherent in the second audiograms (corresponding to one respective ear of a group of persons) may be used in order to decrease uncertainty about the first audiograms (corresponding to the respective other ear of said group of persons). E.g., for the second set of frequencies, the similarities of a user's hearing threshold, measured at one ear, to the second audiograms (given in terms of the weighted mean and/or the respective weighted variance metric) may be used in order to determine initial values for the weighting factors for the first audiograms corresponding to the other ear.
In another embodiment, each of the first audiograms is measured as an in-situ audiogram in the presence of a hearing aid at the respective ear of the corresponding person out of the first plurality, and each of the second audiograms is measured in the absence of a hearing aid at the respective ear of the corresponding person out of the second plurality. This particularly means that the first audiograms, as well as preferably the audiogram to be estimated for the specific user, are in-situ audiograms, measured with a hearing aid located on or at or in the ear of the person corresponding to a respective first audiogram during measurement of the hearing threshold. The second audiograms, in particular, are given by “conventional” audiograms measured, e.g., by an audiometer or with help of a electroencephalography measurement, without any hearing device in or on or at the ear during the measurement. The second plurality of persons may comprise more persons than the first plurality, as it is typically easier to get conventional audiograms, while it is more difficult to get in-situ audiograms. So, this way, the data scarcity for the first, in-situ audiograms (and the statistical drawbacks resulting therefrom) may be compensated for by using also statistical information from the “conventional” second audiograms (which are better available), and in particular, using the intrinsic correlation for the persons comprised both in the first and second plurality. E.g., for the second set of frequencies, the similarities of a user's hearing threshold, measured as a “normal audiogram”, to the second audiograms (given in terms of the weighted mean and/or the respective weighted variance metric) may be used in order to determine initial values for the weighting factors for the first audiograms in order to estimate a user's in-situ audiogram.
The invention furthermore discloses a hearing estimation system comprising a computerized device and a hearing aid, wherein the computerized device is operationally connected to the hearing aid and wherein the computerized device comprises a graphical user interface, a program storage for storing an executable program and a processor for executing said program to perform the method according to one of the preceding claims. The system according to the invention shares the benefits of the method according to the invention. The advantages of the proposed method and of its preferred embodiments can be transferred to the system itself in a straight forward manner. The computerized device may in particular comprise a computer with a CPU and a RAM addressable via said CPU, configured to implement the mathematical operations of the method on the first (and possibly second) audiograms and their respective weighting factors, and further comprising suitable means for measuring a hearing threshold of the specific user, such as, e.g., an audiometer.
The attributes and properties as well as the advantages of the invention which have been described above are now illustrated with help of drawings of an embodiment example. In detail,
Parts and variables corresponding to one another are provided with the same reference numerals in each case of occurrence for all figures.
In
In order to estimate an audiogram for the specific user 1, a first plurality of first audiograms A1j is provided from a database, each of which corresponding to a hearing impaired person Pj out of a first plurality 2 of persons. This means in particular that a database with first audiograms A1j, taken from hearing impaired persons Pj via respective hearing threshold measurements over frequency, are given, said hearing impaired persons Pj amounting to the first plurality 2 of persons.
Furthermore, a corresponding weighting factor wj is assigned to each of the first audiograms A1j, and an initial value lj for each of these weighting factors wj is provided. In other words, a relationship is established between each hearing impaired person Pj, his/her first audiogram A1j, the associated weighting factor wj and the initial value lj for said weighting factor wj. In the following, this direct relationship will be explicitly used.
When no further knowledge on the hearing impaired persons Pj is used, the initial values lj may be provided equal for all weighting factors wj, i.e., wj=1∀j (equivalently, wj=1/N may be used ∀j, being N the total number of first audiograms A1j). However, it is also possible to provide different initial values lj, in dependence on demographic resemblance of a hearing impaired person Pj with the specific user 1. In particular, a higher or lower initial value lj may be provided for weighting factors wj corresponding first audiograms A1j from hearing impaired persons Pj out of the same or out of different age groups, working background groups, residential settlement type etc. The differences to be provided in the initial values lj according to the demographic grouping may be obtained from statistical considerations of the first audiograms A1a with respect to said demographic grouping. For example, the initial value lj for a weighting factor wj may be set to lj=1 for all hearing impaired persons Pj out of the same 10-year age-group as the specific user 1, and lj=0.8 for hearing impaired persons Pj out of a neighboring 10-year age-group, lj=0.6 for the next neighboring 10-year age-group and so on. Furthermore, the first plurality 2 of persons Pj may be chosen out of a multiplicity of persons (not shown) which is a superset to the first plurality 2 of persons Pj, to which respective first audiograms A1j are available, in dependence on demographic similarity of the persons Pj with the specific user 1. Thus, the first plurality of first audiograms A1j for estimating the audiogram of the specific user 1 is chosen out of the multiplicity of the first audiograms A1j each of which corresponding to one out of said multiplicity of persons.
For each frequency fk out of a first set s1 of frequencies, a weighted variance metric vk (wj) for the first audiograms A1j is determined in dependence on the corresponding weighting factors wj. The weighted variance metric vk (wj), in particular, at each frequency fk may be calculated as a weighted sum of the squared deviations of the respective values of the first audiograms A1j from their weighted mean μ, at that frequency:
and N being the total number of first audiograms A1j (i.e., the cardinality of the first plurality 2 of persons). Hereby, the argument wj of vk shall indicate that the weighted variance metric vk depends on the particular values of the weighting factors wj. In case that the initial values lj for the weighting factors wj have been chosen to lj=1∀j, then the weighted variance metric vk yields the normal variance.
Preferably, the first audiograms A1j always show hearing threshold measurements of the underlying hearing impaired persons Pj at the same set of frequencies, i.e., the first set s1 of frequencies fk. However, in case that there are first audiograms A1j which do not have a hearing threshold measurement at one or several frequencies fk out of the first set s1 of frequencies (they may have hearing threshold measurements at alternative frequencies), a linear interpolation or a continuous fit (e.g., a polynomial fit) may be performed on each of these first audiograms A1j, based on the frequencies fk out of the first set s1 of frequencies at which these audiograms do have hearing threshold measurement data.
The weighted variance metric vk is indicated in
Now, a hearing threshold th of the specific user 1 is measured at the first frequency f1, i.e., th (f1). This can be done, e.g., by audiometry or similar methods. In general, a probe sound signal of a known SPL deemed sufficiently high for the specific user 1 hearing said probe sound signal is being presented to the hearing of the specific user 1, and the SPL is decreased until the specific user 1 indicates that he cannot hear the probe sound signal anymore.
Then, for each of the first audiograms A1j, the corresponding weighting factor wj—still being at its respective initial value lj—receives a first update 10. This first update 10 is performed for each weighting factor on the basis of the difference A1j(f1)−th(f1), i.e., in dependence on the difference of the first audiogram A1j (corresponding to the weighting factor wj to be updated), taken at the first frequency f1, and the measured hearing threshold th of the specific user 1 at the first frequency f1. The first update 10 is preferably done in a way that for A1j(f1)=th(f1), the weighting factor wj attains the maximal possible value, and for an increasing difference, A1j(f1)−th (f1), the weighting factor wj decreases more and more, as a result of the first update 10.
After the first update 10, the weighting factors wj are used to generate an estimated audiogram au for the specific user 1, by means of a weighted sum of the first audiograms A1j, as
Note that the estimated audiogram au (fk) as given in eq. (iii) corresponds to the weighted mean μ (fk) as given in eq. (ii) (with the particular exception that the estimated audiogram au is being calculated with the updated weights wj.
In
If this is not the case, then the estimated audiogram au (f1) for the first frequency f1 may also be set to be equal to the measured hearing threshold th (f1) at the first frequency f1 as a design choice, i.e., the value au (fk) as indicated in eq. (iii) is taken only for frequencies fk f1. The estimated audiogram au be used directly for an individual fitting of a hearing aid of the specific user 1.
However, the process shown in
In each of the diagrams 16, 18, 20, the progress of an iteration of the method shown in
In the left diagram 16, only one observation been made at 8 kHz, i.e., the first frequency f1 of
It can be seen that the weighted variance metric vk—as represented by the variance corridor 22—is smallest for the first frequency f1=8 kHz. This is reasonable, as at this frequency there does not exist any uncertainty anymore about the true hearing threshold th of the specific user 1, as it has already been measured. However, at lower frequencies, there still exists a relatively high uncertainty about the validity of the hearing threshold as predicted by the estimated audiogram au.
The weighted variance metric vk (wj), according to the first update of wj, is then used to determine, in an iteration step, a distinguished frequency fa, for which this weighted variance metric is largest. This distinguished frequency fa, for the first update of the weighting factors wj is found at 250 Hz. The next measurement of the hearing threshold th (fd) of the specific user 1 is performed at said distinguished frequency fd, yielding a result of slightly below 80 dB, as it can be seen in diagram 18. By means of said measured hearing threshold th (fd) at the distinguished frequency, a second update is performed on the weighting factors wj, on the basis of the difference A1j(fd)−th(fd) of the corresponding first audiogram A1j and the measured hearing threshold th at the distinguished frequency. The second update is preferably done in a way that for A1j(fd)=th(fd), the weighting factor wj attains the maximal possible value, and for an increasing difference, A1j(fd)−th(fd), the weighting factor wj decreases more and more, as a result of the second update. Preferably, the second update is also performed on the basis of the difference A1j(f1)−th(f1) at the first frequency.
Using the weighting factors wj according to the second update, the update for the estimated audiogram au is calculated (solid line in diagram 18). As it can be seen from the variance corridor 24 (dash-dotted line) in diagram 18, representing the weighted variance metric vk with respect to the update for the estimated audiogram au, the maximum uncertainty is now at 2 kHz, and the variance corridor 24 has a maximum width (i.e., distance to the estimated audiogram au) of slightly more than 20 dB, down from the about 40 dB of the variance corridor 22 in diagram 16. Furthermore, the deviation of the estimated audiogram au from the underlying true audiogram aτr of the specific user 1 has been reduced drastically by the second update, from more than 30 dB at 250 Hz in diagram 16 to about only 5 dB at 1 kHz and 4 kHz in diagram 18.
Diagram 20 shows the estimated audiogram au after a further iteration step, in particular employing a measurement of the hearing threshold th at the next distinguished frequency fd=2 kHz as identified from diagram 18. The next (now third) update is performed on the weighting factors wj, and the update of the estimated audiogram au (solid line in diagram 20) is determined in an analogous way as described above. In diagram 20, said updated estimated audiogram au is displayed along with the corresponding variance corridor 26 representing the weighted variance metric (with respect to the updated estimated audiogram au). The maximum width of the variance corridor 20 is further decreased to about 15 dB at 1 kHz.
Further iteration steps may be performed in a similar way. The iteration may be performed a fixed number of runs, or may be stopped after the weighted variance metric vk (according to the most recent update for the weighting factors wj) falls below a given threshold, or the estimated audiogram au (according to the most recent update for the weighting factors wj) may be visualized and presented to a hearing care professional for decision on whether or not to perform any further iteration steps.
An example for an analytic formula for an iterative updating of the weighting factors wj may be given in dependence on the following function after m iteration steps (m hearing threshold measurements):
The quantity Zj for each first audiogram A1j represents the square-summed deviation from the measured hearing thresholds th (fk) over the frequencies fk. Then, for the case of equal initial values lj=1∀j (or lj=1/N), the m-th update of the weighting factor wj, i.e., after the m-th hearing threshold measurement, may be given by
The relation of eq. (v) is commonly denoted as softmax(−Z) of the vector Z with elements Zj as given by equation (iv).
In case that not all initial values lj are equal, the formula for the m-th update (i.e., after hearing threshold measurements at m different frequencies) may be adjusted to
wherein the denominator m+1 ensures that for each iteration step m, the initial value lj has less contribution to the update of the weighting factor wj.
In
In the embodiment of
Now, the same way as described in
The total variance vtot of the second plurality of second audiograms A2j is now compared to the corresponding weighted variance metric vk of the first plurality of first audiograms A1j at each frequency fk. In case that the difference remains below a given threshold for all frequencies fk, this means that the statistical information of the second audiograms A2j (the “normal” audiograms) is not too different from the statistical information contained in the first audiograms A1j (the in-situ audiograms), so that the determination of the first frequency f1 and/or the update of the weighting factors wj for the first audiograms A1j may be performed also on basis of the second plurality of second audiograms A2j, and in particular on their total variance vtot.
The estimated audiogram au of the specific user is then determined from the first audiograms A1j and the corresponding weighting factors wj for the first audiograms A1j, with the values resulting from the first update 10, according to equation (iii), as described in
Even though the invention has been illustrated and described in detail with help of a preferred embodiment example, the invention is not restricted by this example. Other variations can be derived by a person skilled in the art without leaving the extent of protection of this invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| PA202100378 | Apr 2021 | DK | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2022/059397 | 4/8/2022 | WO |
