Procedure and device for examining hearing ability

Abstract

In a method for examining the auditory capacity of at least one ear of a mammal, growth curves are determined from the measurement of DPOAAs evoked by pairs of excitation signals (f1, f2) for different excitation frequencies f2. First excitation signals with a first excitation frequency f1 and a first sound level L1, and second excitation signals with a second excitation frequency f2 and a second sound level L2, are presented to the ear. Pairs of pulses comprising a first pulse of the first excitation signal and a second pulse of the second excitation signal are presented, and the DPOAAs evoked by these are detected and evaluated. A set of at least two different pairs of pulses with different second excitation frequencies f2 is presented in a block that is repeated several times during a measurement period.

Description

[0001] The present invention relates generally to a method for investigating the hearing ability of at least one ear of a mammal. Based on the measurement of DPOAEs (distortion products of otoacoustic emissions) evoked by pairs of excitation signals (f1, f2), growth curves are determined for different excitation frequencies f2, with the following steps: The ear is presented with first excitation signals with a first excitation frequency f1 and a first sound level L1 and second excitation signals with a second excitation frequency f2 and a second sound level L2; and pulse pairs with a first pulse of the first excitation signal and a second pulse of the second excitation signal are presented in the ear, and the DPOAEs evoked thereby are recorded and evaluated. The inventive method according to claim 1 is further characterized in that a block comprising at least two different pulse pairs with different second excitation frequencies f2 is repeated several times during a measurement period, wherein the duration of the first and second pulses in a pulse pair is 2 to 20 ms and wherein each block of pulse pairs is repeated during a block time T B is presented, which is chosen such that there is a time interval of 30 to 100 ms between the start of a first and a subsequent pulse pair with the same excitation frequency f2, wherein in a block the start of a pulse pair follows the start of the pulse pair immediately preceding it in the block with a time interval T, where T is at least equal to the length of the preceding pulse.

[0002] The generation process of DPOAEs and the measurement and evaluation methods used to date are described, for example, in Dalhoff et al., "Sound and velocity DPOAEs", in HNO 2010, 58: 543-555. Further evidence can also be found there, to which explicit reference is made. Luhashkin & Russell (J. Acoust. Soc. Am. 112(4), 2002; pp. 1561-1568) describe the measurement of DPOAEs in guinea pigs.

[0003] The established methods serve to objectively and quantitatively determine sound processing in a mammalian ear and thus to investigate and subsequently evaluate hearing ability. These methods are based on the measurement of distortion products of otoacoustic emissions (DPOAEs) generated by pairs of excitation signals.

[0004] The measurement results can also be used for adjusting hearing aids in accordance with EP 2 053 877 A1 and DE 199 05 743 A1.

[0005] The auditory system can be viewed as a chain of processing units that are traversed before the more complex perception of hearing arises in the cortex. The first units are the outer ear (auricle and ear canal), the middle ear (ossicles with the footplate forming the boundary to the fluids of the inner ear), and the fluid-filled inner ear. These three units are also called the periphery; they are followed by several neural processing nodes before the signals reach the cortex. The inner ear includes the cochlea, which represents the receptor field for auditory perception and in which sounds are decomposed into their individual frequencies, similar to a Fourier analysis.

[0006] Most hearing loss originates in the inner ear. This includes age-related hearing loss, which on average leads to a 35 dB hearing loss in men aged 60 to 70 and a 25 dB hearing loss in women of the same age group at frequencies above 4 kHz.

[0007] This age-related hearing loss is dominated by an impairment of the so-called cochlear amplifier (the mechanical amplification of the traveling wave in the cochlea) in the inner ear, which in a healthy state, through a complex interplay of electro-mechano-biochemical mechanisms, achieves an amplification of the vibration in the inner ear by a factor of 300 to 1000 before the vibrations are converted into neural signals by the inner hair cells.

[0008] The key element of the cochlear amplifier is the outer hair cells, which function essentially like piezoelectric actuators. Unlike most middle ear injuries, damage to the cochlear amplifier cannot yet be successfully treated.

[0009] The condition of the middle ear can usually be adequately assessed using tympanometry. The condition of the entire auditory system is assessed subjectively through pure-tone audiometry and speech intelligibility tests, and objectively by recording neural excitation.

[0010] If damage to the middle ear component is ruled out, the remaining distinction in the case of hearing loss lies between a neural and a cochlear disorder. Otoacoustic emissions (OAEs) are used to determine this. OAEs are active acoustic emissions from the ear that travel retrogradely, i.e., in the opposite direction to sound perception, via the ossicles and eardrum into the ear canal, where they can be recorded using highly sensitive microphones.

[0011] Two different types of OAEs are distinguished: spontaneous OAEs and evoked OAEs, which are triggered by acoustic stimuli. Spontaneous OAEs (SOAEs) occur in 35 to 50% of healthy ears and are inaudible to the person experiencing them, thus having no significant clinical relevance.

[0012] Evoked otoacoustic emissions (EOAEs) occur during or shortly after acoustic stimulation of the ear. Depending on the type of acoustic stimulus, different subgroups of evoked OAEs are distinguished, including in particular the transiently evoked OAEs (TEOAEs), which are detectable after a short acoustic stimulus, and the distortion product otoacoustic emissions (DPOAEs), which are generated by two simultaneously applied sine tones (f1 and f2).

[0013] The inventive method for determining the state of the cochlea is based on the measurement of DPOAE, i.e. the subgroup of OAE that has been particularly intensively researched.

[0014] DPOAEs are byproducts of a healthy, active inner ear amplifier, which in humans and other mammals amplifies the vibrations generated in the inner ear by an acoustic stimulus by a factor of 100 to 1000 before they are converted into neural signals. The cochlear amplifier uses external energy for this process, which must be supplied by metabolism.

[0015] DPOAEs are generated by simultaneous stimulation with two primary tones, f1 and f2, at excitation levels L1 and L2. These two primary tones, f1 and f2, are so-called "pure tones," meaning pure tones that each contain only one frequency. According to the general Fourier relationship between the time and frequency domains, these tones would have to last infinitely long, otherwise the spectrum would broaden. Since this infinite duration is not achievable, those skilled in the art understand them to be tones that are presented long enough for their spectrum to be sharp. Due to a pronounced nonlinearity in the characteristic curve of the cochlear amplifier, distortion products are generated, some of which are transmitted retrogradely back into the ear canal and can be measured there with suitable instruments. A specific distortion product is evaluated, which preferentially lies at the frequency fdp = 2f1 - f2.Its amplitude allows conclusions to be drawn about the state of the cochlear amplifier, which are valuable in general clinical practice, for example in screening newborns for hearing impairments requiring treatment.

[0016] The amplitude of the DPOAE is typically extracted from the spectrum of the measured signal using Fourier transformation. Since DPOAEs have a very low sound level, which can be significantly below the threshold of hearing, a sufficiently long averaging period is necessary to obtain a certain signal-to-noise ratio and thus a reliable diagnostic result.

[0017] In diagnostic applications, the frequency ratio f2 / f1 is preferably kept constant because in every species there is a frequency ratio at which the DPOAE is strongest and therefore easiest to measure. In humans, this ratio is 1.2, corresponding to a minor third of an octave.

[0018] The main part of the DPOAE signal is produced in the inner ear at the point where f2 is actually represented. This is because the traveling wave of f2 is at its maximum there, and the traveling wave of the f1 tone is already quite strong at this location. At the more apically located point of maximum of the f1 traveling wave, the f2 wave has already completely collapsed. Therefore, the f2 representation site is the one where both tones are relatively strong and are thus processed simultaneously by the highly nonlinear characteristics of the ion channels of the outer hair cells.

[0019] The specialist and also the ENT doctor therefore always associate a DPOAE result with the f2 frequency, i.e. comparing a DPOAE stimulated with f2 = 3 kHz with the audiogram at 3 kHz, although the frequency of the DPOAE itself is almost exactly two thirds below.

[0020] Measuring multiple DPOAEs at a frequency f2 and various sound levels L2, and combining them into a so-called growth curve, provides a more precise indication of the cochlear amplifier's function in the inner ear. For each frequency f2, the so-called threshold can then be determined from the growth curve. This threshold represents the lowest excitation level L2 at which the DPOAE still achieves a given minimum signal-to-noise ratio. This threshold cannot be measured directly, as the measured noise is finite, but must be determined by extrapolation.

[0021] To obtain a diagnostic statement across the entire frequency range, typically 6 to 8 growth functions are measured sequentially.

[0022] The growth curves determined in this way and the threshold values ​​extrapolated from them can then serve as a basis for improved diagnostics and for the adjustment of hearing aids, because the extrapolated threshold values ​​can be seen as a direct statement about the hearing loss, see the DE 199 05 743 A1 mentioned at the beginning.

[0023] The excitation of the DPOAE can be achieved using either continuous or pulsed tones. As previously mentioned, continuous tones are those presented long enough to have a sharp spectrum. In pulsed DPOAE, f1 is introduced as a continuous tone or pulsed, and f2 is also pulsed, with the ratio of L2 to L1 set within a specific range and then L2 being changed stepwise. According to the general Fourier relationship between the time and frequency domains, pulsed tones have a broadened spectrum due to the short pulse. If one of the tones, such as the aforementioned f1, is presented as a continuous tone, this means that the pulsed f2 undergoes a switching-on and switching-off process while f1 is sustained.

[0024] By measuring growth functions using pulsed DPOAE, certain artifacts known as the "two-source problem" in continuous tone DPOAE can be prevented.

[0025] The method is based on two key components: 1) extraction of the so-called nonlinear component of DPOAEs using pulsed stimulation and temporal isolation; 2) measurement of growth functions of pulsed DPOAEs using the extrapolation method of Boege & Janssen, with the important modification of "high-level saturation correction". Both methods are described in Dalhoff et al., "Two-source interference as the major reason for auditory-threshold estimation error based on DPOAE input-output functions in normal-hearing subjects", in Hearing Research 296 (2013), pages 67-82.

[0026] The present invention deals in detail with the improvement of this method.

[0027] The known method has a drawback that prevents its use in routine ENT practice. Estimating the hearing threshold at a single frequency takes approximately 480 seconds with current methods; however, obvious measures to reduce this measurement time to 96 seconds have been discussed. If the standard set of seven frequencies used for clinically assessing hearing ability is to be tested, the measurement time for individuals with normal hearing is already 11 minutes.

[0028] A significant disadvantage of the known method is its slowness, which has therefore led to its previous unsuitability for clinical application, particularly because it is far too time-consuming for examining hearing-impaired individuals. Furthermore, the achievable accuracy in determining the growth curve and extrapolated threshold values ​​is often unsatisfactory.

[0029] This is where the further developed method of fast pulse distortion product otoacoustic emissions (pulse-DPOAE) according to the invention comes in, which aims to create a fast yet accurate method of the type mentioned above.

[0030] According to the invention, this problem is solved in the aforementioned method by presenting a set of at least two different pulse pairs (each pulse pair with an excitation frequency f1 and f2) with different second excitation frequencies f2 (and consequently also different first excitation frequencies f1) in a block that is repeated several times during a measurement period. An example of two different pulse pairs in a block is a first pulse pair with an excitation frequency f2 of 1.5 kHz and an excitation frequency f1 of 1.25 kHz, and a second pulse pair with an excitation frequency f2 of 4 kHz and an excitation frequency f1 of 3.33 kHz. The excitation frequencies f2 and f1 of a pulse pair are preferably linked by a frequency ratio f2 / f1 = 1.2. However, this frequency ratio can also be set to another suitable value between 1.15 and 1.35.Since the focus here is on threshold determination, a frequency ratio f2 / f1 of 1.2 or 1.22 is advantageous, as current knowledge indicates that the frequency ratio is hardly frequency-dependent at low excitation sound pressures. When f2 pulse pairs are mentioned below, this always refers to f1-f2 pulse pairs, where the frequency f1 is determined from f2 via a defined frequency ratio; and when the excitation level L2 is mentioned in this context, it is assumed that the level L1 of the f1 pulse is calculated from L2 according to a predefined rule.

[0031] Pulsed DPOAEs have a significant disadvantage compared to continuous DPOAEs: the measurement of a frequency-sound level combination generally has a low time utilization and therefore a correspondingly lower signal-to-noise ratio for the same measurement time, which is why the use of pulse pairs seems at first glance to even extend the measurement time.

[0032] However, this disadvantage is significantly reduced according to the invention by entangling several measurements in the time-frequency space, i.e., by presenting them alternately with a time delay. Thus, for example, seven frequencies can be stimulated and analyzed within a block with a time delay.

[0033] The problem underlying the invention is completely solved by the claimed method.

[0034] According to the invention, the process is accelerated by parallel and adaptive stimulation and analysis steps. In recent experiments, the inventors measured growth curves for five frequencies f2 in typically 2.5 minutes using the new method. Estimates indicate that growth curves for all seven frequencies can be measured in 2 minutes with this method.

[0035] The inventors have also devised a threshold approximation method that aims to adaptively measure as few points of the growth function as possible during an individual measurement, preferably 3 or 4, in order to save time without losing accuracy.

[0036] For this, the excitation levels L2 must be selected according to the patient and the frequency-dependent condition of their hearing.

[0037] Furthermore, with the new method, the inventors have demonstrated what they consider to be the best way to date to reliably suppress artifacts caused by physiological processes (generated, for example, by heartbeat, breathing, swallowing, coughing, reflexes of the muscle in the middle ear) in a still fast measurement procedure.

[0038] The inventors were thus able to demonstrate for the first time that the new method makes it possible to estimate the threshold of the cochlear state with a standard deviation of only 3 to 4 dB, with individual threshold estimates having no more than 10 dB errors.

[0039] While the typically six to eight growth functions were previously measured sequentially, these measurements are now performed in an interlocking manner according to the invention.

[0040] In principle, the pulse entanglement method eliminates the disadvantage that pulsed signals do not use the entire time of a measurement block and therefore lose SNR.

[0041] For this purpose, the signal-to-noise ratio of the spectrum is determined during the ongoing measurement based on the signal blocks already recorded. The measurement is terminated as soon as the desired signal-to-noise ratio is reached. Since the background noise varies considerably over time due to physiological processes such as heartbeat, respiration, and swallowing, a noise assessment of each individual block is performed during the measurement, and individual blocks are discarded if the noise exceeds a certain empirically determined threshold.

[0042] The invention further relates, according to claim 12, to a device for carrying out the new method, comprising at least one ear probe to be placed in / on the ear with one or two miniature loudspeakers and a receiver (microphone), wherein each miniature loudspeaker is designed to present a first excitation signal with a first excitation frequency f1 and a first sound level L1 and / or a second excitation signal with a second excitation frequency f2 and a second sound level L2, and wherein the receiver is designed to detect and transmit a DPOAE evoked by the first and second excitation signals, further comprising a computer unit which is programmed and configured to present a set of at least two pulse pairs with different second excitation frequencies f2 in a block, which is repeated several times for a measurement period, such that the pulse pairs of a set are entangled, i.e.They will be presented alternately at different times.

[0043] According to the invention, the first and second pulses in a pulse pair are presented simultaneously or with a slight time delay. Normally, the f1 pulse is switched on 3-10 ms earlier, depending on the frequency, and switched off 3-10 ms later, so that the f1 excitation briefly reaches an equilibrium state during the presentation of the f2 pulse. Larger time delays always entail a potential loss of time. However, to maximize time savings, two pulses of equal or similar length for f1 and f2 can be used, time-delayed so that both excitations occur simultaneously at the most diagnostically valuable imaging site for f2 in the cochlea. In this case, the f1 pulse is switched on approximately 0.1-3 ms later, since its transit time to the imaging site of frequency f2, which is located closer to the basal region (towards the footplate), is shorter than for the f2 wave.If this setting is optimally chosen, no effect occurs through the afferent-efferent feedback loop of the medial olivocochlear reflex.

[0044] The excitation frequency f1 is determined from a given excitation frequency f2 according to a given ratio between f1 and f2, as already described in paragraph

[0030] .

[0045] As already described in paragraph

[0030] , the ratio is preferably f2 / f1 = 1.2, because experience has shown that this ratio yields the strongest possible DPOAE. The ratio can be easily adjusted for further optimization depending on f2 and L2; see, for example, Johnson et al. "Influence of primary-level and primary-frequency ratios on human distortion product otoacoustic emissions", in J. Acoust. Soc. Am. 119, 2006, pages 418-428.

[0046] The first and second excitation frequencies are preferably in the range of 250 Hz to 10 kHz.

[0047] According to the invention, the duration of the first and second pulses in a pulse pair is 2 to 20 ms. The pulse with excitation frequency f1 of a pulse pair begins before the pulse with excitation frequency f2 begins and ends after the pulse with excitation frequency f2 has ended; that is, the pulse with excitation frequency f1 is longer than the pulse with excitation frequency f2 of a pulse pair. It is understood that these relationships can also be exactly reversed, namely that the pulse with excitation frequency f2 begins before the pulse with excitation frequency f1 begins and that the pulse with excitation frequency f1 ends before the pulse with excitation frequency f2 ends, so that the pulse with excitation frequency f2 is longer than the pulse with excitation frequency f1 of a pulse pair. The inventors define pulse length as the so-called full width at half maximum (THB or "full width half maximum" (TFWHM).For this pulse length, the time is determined from when the cosine-shaped rising edge has risen to half the equilibrium value ("steady-state") until the corresponding time in the cut-off edge.

[0048] This pulse length results from the preferred pulse shape, according to which the pulses have a cosinusoidal rise of 0.1 to 4 ms length, a steady state with level L2 or L1, which is 2 to 12 ms long, followed by another cosinusoidal section.

[0049] In a preferred arrangement, the total length of a pulse is measured such that it reaches its full amplitude while the response of the first contribution to the DPOAE, i.e., the nonlinear contribution, begins but is already in the decay process when the response of the second contribution to the DPOAE begins.

[0050] This approach achieves two goals: minimizing the time required to measure the first (nonlinear) contribution of a DPOAE and cleanly separating it from the second contribution. Possible pulse shapes are described in Whitehead et al., "Visualization of the onset of distortion-product otoacoustic emissions, and measurement of their latency," in J. Acoust. Soc. Am. 100 (3), 1996, pages 1663–1679, and Zelle et al., "Extraction of otoacoustic distortion product sources using pulse basis functions," in Acoust. Soc. Am. 134, 2013, pages EL64–69.

[0051] To effectively suppress the primary tones (f1 and f2), in addition to usual filter methods, the primary-tone-phase-variation method of Whitehead et al., 1996, ibid., is preferably used.

[0052] In a block, the start of a pulse pair follows the start of the immediately preceding pulse pair in the block with a time interval T (T SLOT or TS ), where T is generally at least equal to the length of the preceding pulse, i.e., > 10 ms. This measurement time reserved for a pulse pair in a block is subsequently also referred to as a slot. It should be noted that slots do not overlap; rather, one slot follows the other when the preceding slot has ended. As previously explained, the slot length (TS ) is more than 10 ms.

[0053] The advantage here is that the second pulse pair is only presented once the DPOAE evoked by the first pulse pair has sufficiently decayed (to approximately 1 to 10% of the initial value). This prevents any noticeable interference in the measurement of the levels (Ldp) of the individual DPOAEs, which, compared to the sound levels L1 and L2 of the excitation pulses, exhibit only a very low sound pressure level (SPL). Furthermore, the increased time interval between the presentation of pulse pairs with the same excitation frequencies f1 and f2 allows sufficient recovery time for the DPOAE triggered in the previous measurement block to completely decay.

[0054] For example, if a measurement with four pulse pairs is performed and a slot length T (TS) of 40 ms is chosen for each pulse pair in a block, then for each DPOAE of an excitation frequency f2, a recovery time of 3 x 40 ms, or 120 ms, results before the measurement is repeated at the same excitation frequency f2. Within a total block time (TB) of 160 ms (4 x 40 ms = 4 x TS), DPOAEs for four second excitation frequencies f2 can be measured in this example without them influencing each other or due to insufficient decay time. In this example, the block therefore has four slots, each with a slot length TS of 40 ms, where the current occupancy of the slots with pulse pairs of different second excitation frequencies f2 is referred to as a panel.

[0055] The risk of mutual interference between pulse pairs and the DPOAEs they evoke within a block exists because, when stimuli are presented simultaneously, traveling waves can drive the nonlinear transmission in the inner ear into saturation, a phenomenon known as suppression or masking. This distorts the measurement, particularly of simultaneously presented higher frequencies.

[0056] Furthermore, it is preferred that in a block the second excitation frequencies f2 of two immediately consecutive pulse pairs, i.e., in two consecutive slots, are at least one octave apart. The first excitation frequencies f1 are selected as described in paragraph

[0030] .

[0057] Here too, it is advantageous that there are no noticeable interferences during the measurement of the individual DPOAEs, because care is taken to ensure that the frequency difference between the excitation frequencies f1, f2, and thus also between the frequencies fdp of the respective evoked DPOAEs, is sufficiently large. It has been shown that a frequency difference of one octave is sufficient.

[0058] A preferred set (preferred panel) consists of the excitation frequencies f2 = 1 kHz, f2 = 3 kHz, f2 = 1.5 kHz, f2 = 6 kHz, which are presented as a panel in a block, repeated in that order. A panel is to be understood as the frequency-time pattern of the excitation frequencies in a block.

[0059] Another preferred set (preferred panel) consists of the excitation frequencies f2 = 2 kHz, f2 = 4 kHz, f2 = 1.5 kHz, f2 = 3 kHz, which are presented as a panel in a block, repeated in this order. Because the measurement at 1 kHz typically requires twice as much time to achieve a certain signal-to-noise ratio as the measurement at the other (higher) frequencies, a shorter measurement time is typically required for this set to achieve a desired signal-to-noise ratio (SNR).

[0060] In general, it is preferred to average the measured sound levels Ldp of the DPOAE for pulse pairs of the same first and second excitation frequencies f1, f2 during the measurement period.

[0061] It is advantageous here that the measurements with the blocks of the set of pulse pairs are repeated over a certain period, the measurement duration, because the SNR is improved when the DPOAE for pulse pairs of the same excitation frequencies f1, f2 are averaged.

[0062] Each block of pulse pairs is presented during a block time TB, which is chosen such that there is a time interval of 30 to 100 ms, preferably at least 70 ms, between the start of a first and a subsequent pulse pair with the same excitation frequency f2. As already explained, the block time TB is the sum of the slot lengths TS of a block.

[0063] According to the inventors' findings, this time interval is sufficient for a pulse to be sufficiently pronounced when it is repeated.

[0064] In general, it is preferred to check at the beginning of the measurements whether the frequency fdp of one of the DPOAEs interferes with a spontaneous emission (SOAE).

[0065] Here, it is advantageous that artifacts and sources of interference are detected at the beginning of the measurement. If this is the case, either the block time TB or the duration TS for one or all slots can be adjusted so that the decay time of the DPOAE is extended sufficiently for its level to fall below a certain threshold before the next pulse pair is presented, or the frequency f2 is shifted to establish a minimum distance to the SOAE.

[0066] It is preferred that, at the beginning of a measurement, a DPOAE is measured for a pulse pair with a first excitation frequency f1 and a first sound level L1, and a second excitation frequency f2 and a second sound level L2. If no DPOAE is measurable, the sound level L2 (and the sound level L1) is incrementally increased until either the maximum output sound level L2 (L1) is reached or a DPOAE is measured. In this simple way, the presence of interfering SOAEs can be detected and, if necessary, compensated for.

[0067] Furthermore, it is preferable if the sound levels of the pulse pairs within a block are chosen to be similar, since the risk of mutual suppression increases with increasing level difference.

[0068] In a simple implementation, a fixed threshold can be defined for the maximum level difference between the L2 sound levels of, for example, four pulse pairs in a block, which in a preferred implementation lies between 5 and 15 dB.

[0069] On the one hand, it is preferred if the sequence of pulse pairs and the time interval between two successive pulse pairs, i.e. the slot time (TS), are constant in a block.

[0070] In this block-based approach, as many blocks as necessary are measured and averaged until the desired SNR is achieved for each excitation frequency in the set (panel).

[0071] On the other hand, it is preferred if, upon reaching the desired SNR for an excitation frequency f2, the remaining pulse pairs planned for this excitation frequency f2 and consequently their averaging are skipped.

[0072] The advantage here is that measurements are taken with the remaining pulse pairs in shortened blocks, i.e. with fewer slots, which further reduces the measurement time.

[0073] Furthermore, it is preferred to select at least two sets with pulse pairs that differ at least partially with respect to the second excitation frequency f2, whereby the blocks of the individual sets are presented sequentially and the DPOAEs are measured and averaged. The sets are thus processed one after the other.

[0074] In this block-flexible method with a fixed pulse arrangement, for example, seven pulse pairs with different excitation frequencies f2 are arranged in such a way that, according to general experience, they are all allocated approximately the averaging time required to achieve a certain signal-to-noise ratio.

[0075] The first set contains approximately four pulse pairs, although a total of seven are to be measured. If all seven pulse pairs were combined into one block, the recovery time for each individual frequency would be excessively long, e.g., 6 x 40 ms = 240 ms.

[0076] In order to obtain a sensible compromise between the shortest possible measurement time and sufficient time for each DPOAE to decay, according to the invention several sets are presented one after the other, onto which the, for example, seven pulse pairs are distributed in such a way that pulse pairs with low excitation frequencies f2 (and thus also low excitation frequencies f1) occur in several sets.

[0077] Because background noise (1 / f noise) increases at excessively low excitation frequencies f2, this assignment leads to DPOAEs being averaged more frequently for pulse pairs with lower excitation frequencies f2 than for those with higher excitation frequencies f2. This further reduces the measurement time.

[0078] However, if one of the excitation frequencies f2 in a patient only leads to a very weak DPOAE, for example at f2 = 3 kHz, then despite the good noise background at this excitation frequency, the averaging must be much longer than usual, which is why the other excitation frequencies f2, which are also presented in the measured set, are stimulated for an unnecessarily long time, which could again extend the measurement time in such situations.

[0079] In further training, it is therefore preferred if it is continuously checked for each pulse pair whether a desired SNR is achieved, and if, in further measurement, the pulse pairs for this excitation frequency f2 are eliminated and the remaining pulse pairs are possibly redistributed to the blocks.

[0080] In this block-flexible method with free pulse pair arrangement, the measurement lengths for the individual pulse pairs are no longer fixed relative to each other. While some of the previously described method variants also only averaged until the desired SNR was achieved for each excitation frequency f2, this meant that all pulse pairs had to wait for the last pulse pair, resulting in all but one excitation frequency f2 having a higher SNR than required, at the cost of excessively long measurement times.

[0081] According to the invention, it is therefore continuously checked for each pulse pair whether the SNR has been reached; as soon as this is the case, it is checked whether another pulse pair is still incomplete. In this way, the completed pulse pairs are successively eliminated from the measurement, and only the remaining pulse pairs are presented.

[0082] The process now checks, firstly, whether the octave interval between two consecutive pulse pairs is maintained. If this is no longer the case, pulse pairs may no longer be processed in the blocks to which they were originally assigned, but in other (newly defined) blocks. Secondly, it checks whether the required time interval T (corresponding to the time for a slot (TS) plus the necessary decay time (Tdec)) is maintained between pulse pairs with the same excitation frequency f2.

[0083] Preferably, the DPOAE is measured and averaged for all excitation frequencies f2 contained in the set(s) at a sound level L2 assigned to each excitation frequency, and then at least one further measurement is carried out at new sound levels L2, wherein preferably in a threshold approximation method the new sound level L2 for each further measurement is determined from the measured DPOAE for each excitation frequency f2.

[0084] This procedure is repeated until a growth curve can be determined for each excitation frequency f2 from measured values ​​of the sound levels of the DPOAE for 3 to 4 different sound levels L2, from which the respective threshold values ​​are then determined.

[0085] Typically, based on the first measured DPOAE, the next lower sound level L2 is calculated for each excitation frequency f2. If the starting point was an L2 of 45–55 dB SPL, it is highly likely that the next lower sound level will be between 35 and 50 dB SPL. If the seven calculated new sound levels satisfy the specified condition for the maximum sound level spacing between two pulse pairs, the previously used procedure can be applied again to determine a mean sound level Ldp of the DPOAE for each growth curve at the position of the new sound level L2.

[0086] In this context, it is preferred if the growth curve is determined from at least three values ​​for sound level Ldp of the DPOAE determined at three different sound levels L2 but the same excitation frequency f2, wherein the at least three sound levels L2 include an upper sound level L2, with the help of which a lower sound level L2 is determined, wherein the third, middle sound level is determined with the help of the upper and the lower sound level, and preferably with the help of formulas (1) to (14) from the following section 2.2.3. A preliminary lower sound level L2 is determined from the upper sound level L2 and population data, and then the mean sound level L2 is determined from the upper sound level L2 and the preliminary lower sound level, which is preferably midway between the upper and the preliminary lower sound level L2, and furthermore, preferably using formulas (1) to (14), a final lower sound level L2 is determined from the upper sound level L2 and the mean sound level L2.

[0087] This iterative procedure first selects the upper sound level L2, then estimates a lower sound level using population data and the upper sound level, then determines the mean sound level between the upper and the preliminary lower sound level, and then determines the lower, threshold-adjusted sound level again from the upper and the mean sound levels using the individual slope that can now be determined.

[0088] However, it is possible that the sound level condition is not met for one or more excitation frequencies f2. It is also possible, in principle, that the threshold approximation method will result in a lower sound level L2 being set for some excitation frequencies f2, while no DPOAE could be measured for other excitation frequencies f2, and therefore a higher sound level L2 must now be used for excitation.

[0089] The blocks must then be adjusted accordingly by defining new panels, whereby it may be necessary to carry out individual measurements for certain excitation frequencies f2.

[0090] If, for example, a patient only shows hearing loss at 6 kHz, nothing can be measured after the initial presentation at a starting level of, for example, 45 dB SPL.

[0091] The excitation frequency f2 = 6 kHz is not presented in a block-flexible method with a fixed pulse arrangement. Once all sound levels L2 of the remaining excitation frequencies f2 have been processed, up to three individual measurements with a block length of 70 to 120 ms follow in order to determine a threshold value even at f2 = 6 kHz.

[0092] In contrast, a block-flexible method with free pulse arrangement allows for more flexible operation. Here, pending pulse pairs are continuously checked for compatibility with regard to sound level and time interval. Furthermore, it is checked whether pending pulse pairs can be moved to available slots. This can reduce the number of individual measurements in many cases.

[0093] The new procedure and device can be used commercially in newborn hearing screening, by pediatricians in UX examinations, in ENT clinics, and by ENT doctors, in each case for the examination of hearing ability.

[0094] They can also be used by hearing aid acousticians for the fitting of hearing aids, which can now be done automatically rather than iteratively, and by patients who can perform an automated hearing test at home to check, for example, whether they need to see a doctor or clinic, or whether their hearing aid needs to be readjusted.

[0095] There are also plans to integrate the new device into hearing aids, where it in situ to automatically adjust the hearing aid equipped with it to a change in the hearing ability of the wearer of the hearing aid.

[0096] Further advantages are outlined in the description and the accompanying drawing.

[0097] Exemplary embodiments of the invention are explained in more detail in the following description with reference to the accompanying drawings. These show: Fig. 1 a schematic diagram of a device with which the new method is carried out; Fig. 2 a schematic growth curve as it can be recorded and extrapolated with the new method; Fig. 3 for the block-rigid time-frequency pulse entanglement method with fixed pulse pair arrangement the envelopes for four different f2 excitation pulses with frequencies of 1.5; 4; 2; 3 kHz; Fig. 3b for the block-rigid time-frequency pulse entanglement method with fixed pulse arrangement the envelopes for four different pulse pairs, each with four different excitation frequencies f2 and f1, with frequencies f1 1.25; 1.67; 2.5; 3.33 kHz; and frequencies f2 1.5; 2; 3; 4 kHz. Fig. 4 for the block-rigid time-frequency-pulse entanglement method with fixed pulse pair arrangement, the division of a panel of four f2 pulse pairs into four slots in one block; Fig.Fig. 5 for the block-rigid time-frequency-pulse entanglement method with fixed pulse pair arrangement, the distribution of a total of seven f2 pulse pairs across two panels and two blocks, each with four slots; Fig. 6 for the block-flexible time-frequency-pulse entanglement method with fixed pulse pair arrangement, the distribution of a total of seven f2 pulse pairs across three panels and three blocks, each with four slots; Fig. 7 for the block-flexible time-frequency-pulse entanglement method with free pulse pair arrangement, the distribution of a total of seven f2 pulse pairs across four panels and four blocks with a variable number of slots, under the simplified assumption of uniform measurement times for selected f2 pulse pairs; and Fig. 8 for the block-flexible time-frequency-pulse entanglement method with free pulse pair arrangement, the distribution of a total of seven f2 pulse pairs across six panels and six blocks with a variable number of slots, without the simplified assumption of uniform measurement times for selected f2 pulse pairs.

[0098] A device 10 to be used according to the invention is shown schematically in Fig. 1 shown, it typically includes at least one ear probe 11, which is connected via cable 12, 14 to a sound card or another A / D card in a computing unit 15, here a computer.

[0099] According to the aforementioned DE 199 05 743 A1, a wireless connection can be provided between headphones with two ear probes and the computer to avoid interference that could be caused by the cable. Because of the two ear probes, binaural measurements are possible.

[0100] Each ear probe 11 contains at least one highly linear loudspeaker 16, which emits the two excitation signals f1, L1 and f2, L2, and at least one microphone 17, which measures the DPOAEs, i.e., their sound pressure levels Ldp at a frequency fdp = 2 f1 - f2 set using f2. However, each ear probe 11 often contains two loudspeakers 16 to prevent technical distortions from occurring during stimulation with the two tones f1, f2, which would be difficult to distinguish from the physiological DPOAEs.

[0101] The computer can be externally mounted or integrated into the ear probe 11. It is designed to generate the pulse pairs (f1, f2) used according to the invention and to adapt them adaptively during the measurement. It can store the pulse pairs (f1, f2) and measured DPOAEs (Ldp, fdp) for later analysis and make them available for retrieval.

[0102] Alternatively, the computer can also perform the analyses in real time.

[0103] The measurement results in so-called growth curves for the selected excitation frequencies f2, from which the computer then determines the respective threshold values. These thresholds represent an objective assessment of hearing ability and can be used for various purposes. Importantly, the new method enables a very rapid determination of the growth curves, which promotes the acceptance of the procedure.

[0104] A basic growth curve 21 is in Fig. 2 The measured level Ldp of the DPOAE for an excitation frequency f2 is shown. The threshold value 23 is extrapolated from the seven measured values ​​22 for the DPOAE.

[0105] To adapt the signals f1, f2 output by the computer to the loudspeaker 16 built into the ear probe 11, a power amplifier and / or an impedance matching and frequency response correction can be provided, which is implemented in the form of a passive or active electronic circuit.

[0106] To adapt the electrical signals generated by microphone 17 to the computer interface, a preamplifier and a frequency response correction can be provided, which is also implemented in the form of a passive or active electronic circuit.

[0107] If not all of the described processing stages are located in the ear probe 11, then either a wired or a wireless connection is provided. If all processing stages are located in the ear probe 11, a wireless connection is provided either for transferring the measurement data after the measurement or in real time to a playback device.

[0108] The data can also be stored as clinical data in clinical information systems.

[0109] The devices described above (10) are used to examine hearing ability by pediatricians, in ENT clinics, by ENT physicians, by hearing aid acousticians, and by patients at home. They can also be integrated directly into hearing aids to adjust them, so to speak, during operation to changes in the wearer's hearing ability.

[0110] In Fig. 1 A hearing aid 20 is schematically indicated in which the device 10 is arranged.

[0111] The method to be carried out with these devices was initially described in its individual necessary and preferred steps. Below, a complete description of preferred embodiments is given, using slightly modified nomenclature in some cases. 1. Introduction

[0112] A time-optimal method for measuring DPOAE growth functions must consider different levels of measurement acquisition and analysis, which can be subdivided as follows: 1.1) Measurement of a single DPOAE

[0113] This requires defining a minimum achievable signal-to-noise ratio, defining how signal and noise are to be calculated, a suitable artifact suppression method, and usually a maximum measurement time after which the measurement is terminated as unsuccessful. 1.2) Threshold approximation

[0114] When measuring growth functions, the question arises as to at which excitation levels, in what order, and how many individual measurements should be recorded. In the case of extrapolation methods, it has been shown that often only the three points closest to the threshold of a growth function contribute to an accurate estimation of the extrapolated threshold, or that the estimation accuracy increases when other points are omitted. This leads to the idea of ​​selectively measuring these three points in order to save the time required for the remaining points.

[0115] As will be shown, the measurement point closest to the threshold is the most critical. If its excitation level is chosen too low, either excessively long averaging is required, or it may even be impossible to determine the value with the required certainty after the maximum individual measurement time has elapsed. If it is chosen too high, the extrapolation error increases because the last measurement point is further away from the threshold being estimated.

[0116] Therefore, a procedure is provided which, from a first valid single measurement value, which should preferably be the third and most distant measurement point, determines the two following measurement points according to optimal criteria, whereby with each additional measurement point the calculation of the optimal next excitation level is to be adaptively refined.

[0117] A separate consideration is given to the question of how to stimulate the first and most distant individual measurement value. This involves, in a sense, a global strategy that can vary depending on the application.

[0118] Generally, it is advisable to excite the first measurement at approximately L2 = 45 dB SPL, as experience has shown that above this value, growth functions in individuals with normal hearing already saturate, i.e., produce values ​​that are not helpful for an optimal threshold estimation. For example, when measuring newborns as part of a screening test, this would be a good choice because the vast majority of newborns have normal hearing, and thus the average measurement time for the screening test would likely be minimized, since in most of the population only three individual measurements (per frequency) need to be obtained, and the first measurement point furthest from the threshold is rarely attempted unsuccessfully.

[0119] In an application for patients seeking treatment for hearing loss, growth functions corresponding to a normal-hearing population are much less common, while the number of cases where the maximum measurement time for a single measurement is consumed at L2 = 45 dB SPL without obtaining a valid value can already be considerable.

[0120] This area should also include methods that reposition a growth function when optimal excitation is not available, and those in which, in multi-frequency measurements (i.e., pulse entanglement methods with multiple frequencies), growth functions are adjusted depending on results at other frequencies. 1.3) Temporal pulse entanglement method

[0121] As a rule, the condition of the inner ear should not be recorded at just one frequency, but at a suitable number of frequencies, typically in octave or half-octave steps in the range 1 kHz ≤ f2 ≤ 8 kHz. This means that several growth functions must be measured.

[0122] Pulsed DPOAEs have a significant disadvantage compared to continuous DPOAEs: the measurement of a frequency-sound level combination generally has a low duty factor and therefore a correspondingly lower signal-to-noise ratio for the same measurement time.

[0123] However, this disadvantage is significantly reduced according to the invention by entangling several measurements in the time-frequency space, i.e., by presenting them alternately with a time delay. Thus, within a sufficiently frequently repeated block, for example, 7 frequencies can be stimulated and analyzed with a time delay.

[0124] The simplest implementation is a fixed arrangement of excitation pulses within a block; this is called the block-rigid method. When arranging the pulse pairs within a block, the frequency and time sequence must be chosen such that the signals interfere with each other only to a negligible extent.

[0125] This method works almost perfectly, especially when the noise background and the occurrence of artifacts are similar at all frequencies used.

[0126] Further optimization is achieved when the stimulating pulse pairs are arranged in a block-flexible rather than a rigid block configuration. This leads to additional time savings if the measurement has already achieved a sufficient signal-to-noise ratio with one frequency-level combination, while another still requires significantly more time.

[0127] In this case, in the block-flexible pulse entanglement method, before the end of one of the individual measurements currently occurring in the block, one or more pulse pairs that have already achieved the required signal-to-noise ratio are replaced by those with a different frequency or level that are still in the measurement.

[0128] For this purpose, a set of rules is provided that defines a minimum distance in the time-frequency level space. Specifically: 2.1) Measurement of a single DPOAE

[0129] Suitable methods for determining criteria for including measured values ​​– here the DPOAE – in averaging, and for calculating an average value that exhibits a predetermined signal-to-noise ratio (SNR), are well known from the prior art; see, for example, Müller and Specht, "Sorted averaging – principle and application to auditory brainstem responses", in Scand. Audiol. 1999, 28: 145-9.

[0130] However, the inventors have found that the decay time of pulse responses can pose a problem for the approach of reducing measurement time by shortening the block length.

[0131] In typical measurements in the frequency range around f2 = 2 kHz, a block length of T = 70 ms is sufficient to ensure that the pulse response within a block has decayed sufficiently that it does not generate any significant measurement errors in the following block due to interference with the pulse response to the renewed stimulus by another pulse pair.

[0132] However, this does not apply if the frequency of the sought-after DPOAE, usually fdp = 2f1 - f2, is close to that of a spontaneous otoacoustic emission (SOAE).

[0133] To limit measurement errors resulting from this problem, the inventors use various approaches: A) Adjusting the block length to the decay time of the pulse below a certain level. This method has the disadvantage that a considerable increase in measurement time may be necessary, and that a certain amount of averaging, with its associated time expenditure, is required to determine the decay time. Therefore, this method is only useful if the measurement is required at a precisely defined frequency. B) In clinical practice, however, measurement at a precisely defined frequency is not necessary. A sufficient picture of the cochlea's condition is generally obtained by measuring at octave frequencies, or, for higher demands, at half-octave frequencies, for example, at f2 = 0.75; 1; 1.5; 2; 3; 4; 6; 8 kHz.

[0134] In such applications, measurements at a frequency f2 that is 50 to 150 Hz away from the target frequencies should be perfectly adequate. Therefore, the existence of SOAEs should be verified, either a) in an a priori measurement, the evaluation of which is used to shift the actual frequency f2 of the stimulus in order to obtain a sufficient safety margin between SOAEs and fdp, or b) by detecting the problem of excessively long decay times when measuring DPOAEs at their target frequencies f2, using an algorithm that runs during the measurement, and then repeating the measurement with shifted stimulus frequencies f2 if necessary.

[0135] The choice between these two options can be determined by the probability of SOAEs occurring in the given application. If SOAEs are only very rarely expected in an application involving individuals with severe hearing loss, one would want to save the time spent measuring SOAEs (e.g., hearing screening in older adults). In contrast, SOAEs are almost certainly present in newborns, and therefore the probability that they will distort the DPOAE measurement is higher.

[0136] Regarding A): A SOAE measurement typically takes 40 seconds to achieve a noise background of -30 dB SPL at f = 2 kHz. The required frequency offset can be adjusted depending on the strength of the detected SOAE.

[0137] Regarding B), the optimized averaging algorithm is a suitable option here. It predicts that when blocks with unusually high noise are excluded, the noise will follow a clear pattern. If it does not, or if practically every block is excluded after a certain time, this indicates that the chosen noise measure no longer reflects the assumption of a random, uncorrelated process.

[0138] If the noise measure refers, for example, to the amplitude of the envelope of the time signal immediately before a new pulse response is expected, the averaging process may function correctly for a certain number of blocks. However, once the noise amplitude has decreased to the value of the decayed pulse response to the preceding stimulus pulse, the amplitude will no longer decrease through averaging if the decaying pulse response is highly reproducible and thus exhibits a high correlation.

[0139] A scheme for starting the measurement of a growth function stipulates that the measurement begins with a starting level L2 for the second primary tone at f2. The level L1 for the first primary tone f1 is determined according to a standard curve for the relevant frequency, e.g., using the level shear as described by P. Kummer et al., "The level and growth behavior of the 2f1-f2 distortion product otoacoustic emission and its relationship to auditory sensitivity in normal hearing and cochlear hearing loss", in J. Acoust. Soc. Am., 103(6):3431-3444, 06 1998.

[0140] If no DPOAE can be measured, the level L2 of the second primary tone is increased by ΔL2 until the maximum achievable primary tone level is reached. If no DPOAE is present even then, an attempt is made to optimize the level of the first primary tone, L1.

[0141] If a DPOAE is measured, the threshold approximation procedure described below follows. 2.2) Threshold approximation 2.2.1 Determining the first measuring point

[0142] Since experience has shown that DPOAE growth functions in individuals with normal hearing tend to saturate relatively early, the first measurement should be taken at the highest excitation level at which saturation is not typically expected. This measured point would then still be usable for an individual with normal hearing and thus measured without any time loss. It would be expediently the point on the growth function furthest from the threshold.

[0143] Based on previous experience, an excitation level of L2 = 45 dB SPL is recommended. If this point is chosen too low, the number of cases in which the maximum permissible measurement time is reached without obtaining a usable individual measurement value increases.

[0144] Here, the strategy can optionally be varied as follows: The maximum permissible measurement time is reduced for the first measurement point. This results in less time being lost if the DPOAE threshold is above L2. If it is only slightly below, the acquisition of data for the same measurement point can be resumed in a later step when L2 is to be measured with a longer measurement time. This latter method utilizes the advantage of the comparatively quickly acquired information from the upper points in the growth function to optimally adjust the level of the lowest point of the growth function in terms of time, thus achieving maximum time efficiency. 2.2.2 Determining the second measuring point

[0145] The following analysis assumes the measurement of three pulse pairs for a given f2, i.e., three different sound levels L2 for every second excitation frequency f2. This corresponds to the minimum number of points to which the extrapolation method with quality control (determination of the correlation coefficient and the standard deviation of the estimated threshold) can be applied.

[0146] Measurements in individuals with normal hearing show a high correlation between the slope of the growth function and the amplitude of the DPOAE at a medium excitation level, approximately L2 = 45 dB SPL. It is therefore advisable to select the second excitation level so that it lies midway between the first and third excitation levels. For this purpose, the slope of the growth function is estimated using population means (see Dalhoff et al., 2013, Hearing Researches, Vol. 296, Table 2, p. 77), which are best determined as a function of frequency.

[0147] Using the slope, the time-optimal excitation level L2 for the last measurement point closest to the threshold can then be determined. Section 2.2.3 describes a procedure for this, in which the aforementioned population mean is substituted for m in equation (2). Based on the first (farthest) measurement point and population data, the measurement point closest to the threshold (third) is then estimated. The second measurement point is then placed midway between the first and third measurement points. 2.2.3 Determining the last measurement point at the lowest excitation level L2

[0148] It is assumed that two or more measurement points of a growth function are already available at higher excitation levels L2. According to section 2.2.2, for example, the two upper measurement points are available, so that the third (nearest threshold) measurement point can now be determined again, but this time with the individually determined slope.

[0149] The task is to solve how to choose the last and lowest excitation level L2 in such a way that a minimal estimation error is obtained in the final extrapolation to the threshold of the growth function with the least possible time expenditure.

[0150] Two questions need to be answered. First: What influence does a measurement error of the last measurement point closest to the threshold, resulting from a limited averaging time, have on the regression and thus on the estimation error of the extrapolated threshold? It is clear in principle that a measurement point close to the threshold will lead to a smaller extrapolation error for the same measurement error. Secondly:The final point of the growth function must be measured with a certain noise margin to be considered valid, and this means that the closer it is chosen to the threshold of the growth function, the more the required averaging time will increase. From these two considerations, a clear optimality criterion will emerge.

[0151] The points of the growth function are numbered from the lowest to the highest excitation level, i.e., contrary to the temporal measurement sequence. Consequently, we are looking for the point P 1 with the excitation level x 1, while the points P 2 , ..., P n already available. Based on the available data, a preliminary regression line (or, in the case of n = 3 a connecting line) can be determined, with y ( x ) = m e,n - 1 x + b e,n - 1 , where e represents estimatedIt should be standing, and n - 1 represents the number of points used for the estimate.

[0152] The following assumes that y ( x ) describes the actual growth function without any error, while the newly added measurement point introduces a measurement error due to the finite averaging time. N exhibits 1, so that the measured (or estimated from the measurement) value has an amplitude of y 1 ,e = y 1 + N 1 will be shown if y 1 is the true amplitude.

[0153] The slope of the regression line resulting from the error-prone measurement of P As can be seen from 1, it follows from: m e = xy e ¯ − xy e ¯ x 2 ¯ − x ¯ 2

[0154] The line above the variables signifies averaging. Therefore, Δ m = m e − m = xy e ¯ − xy e ¯ x 2 ¯ − x ¯ 2 − m

[0155] The y-intercept is calculated as follows: b e = y e ¯ − m e x ¯ and Δ b = be - b will Δ b = N 1 n − Δ m x ¯

[0156] The extrapolation error Δ resulting from the measurement error x edpt We can find it using the quotient rule, provided the error is not too large: Δ x edpt = b Δ m − m Δ b m 2

[0157] The following follows: nm Δ x edpt = N 1 x 1 − x ¯ x ¯ − x edpt − β β with the abbreviation β = x 2 ¯ − x ¯ 2

[0158] After separating the desired point x 1 from the means, we find N 1 , reg = nmΔx edpt x 1 2 n − 1 n − 2 γx 1 n + δ − γ 2 n x 1 x edpt 1 − n + γ + γx edpt + δ with γ = ∑ 2 n x i and δ = ∑ 2 n x i 2 .

[0159] The suffix in the index denotes N 1, reg , This refers to the acceptable noise level for the regression criterion. Therefore, the permissible noise at the point is calculated from this equation. P 1, if a maximum extrapolation error Δ x edpt is predetermined.

[0160] The requirement for the signal-to-noise ratio criterion is: N 1 , snr = m x 1 − x edpt SNR

[0161] The signal-to-noise ratio is SNRto be understood as a linear ratio of the amplitudes. From the two noise requirements defined in equations (8) and (9), the required averaging time can be determined if a noise amplitude density Ṅ It is known: τ = N ˙ N 1 2

[0162] Since both conditions must be met simultaneously, the longer measurement time must be observed. Equating the two noise criteria leads to a quadratic equation with the following solution: x 1 = − p 2 − p 2 4 − q with p = − 2 Δ x edpt SNRγ + δ − x edpt 2 1 − n α q = Δ x edpt SNR nδ − γ + x edpt γx edpt + δ α α = Δ x edpt SNR + x edpt n − 1 − γ

[0163] In practice, it may be necessary to deviate from the recommendation. For example, it is known that in people with normal hearing, the growth function is impaired below a level of approximately... L 2 = 25 dB SPLThe curve no longer follows the ideal linear progression in the semilogarithmic representation, since, strictly speaking, there is no distortion product threshold. If an excitation level in this range is recommended based on the method described above, the estimation accuracy is systematically degraded. For example, a minimum value for L 2 1 be determined.

[0164] Alternatively, in a simplified procedure, a set of L2 sound levels can be stored in the computer, which is used for each excitation frequency f2. It is also possible to store a separate set of L2 sound levels for each excitation frequency f2. 2.3) Temporal pulse entanglement method 2.3.1 Time-Frequency-Pulse Entanglement Method 2.3.1.1 Block-rigid time-frequency-pulse entanglement method

[0165] We assume that in one block a panel with several pulse pairs of different frequencies f1, f2 and also levels L1, L2 is presented, and yet from, for example, 4 such blocks, so-called ensembles are formed using the PTPV method (see, for example, Zelle et al., "Extraction of otoacoustic distortion product sources using pulse basis functions", in J.Acoust.Soc.Am., 134(1):EL64-EL69, 07 20139), which enable the extraction of the time signal of a desired distortion component, e.g., at fdp = 2f1 - f2.

[0166] We call this presentation mode block-rigid when the sequence of pulses within a block is predetermined or unchanging during the measurement.

[0167] One possible approach is to arrange n pulse pairs with different excitation frequencies f2(i), i = 1, 2, 3, ...n, in a time-shifted manner within a block. The block length is chosen, for example, to be T (or TB) = 160 ms, and the start times of the pulse pairs are evenly distributed across the block, in this example with n = 4 and f2 = 1.5, 3, 2, 4 kHz. The main focus is then on extracting the distortion component at fdp = 2f1-f2.

[0168] In Fig. 3 For the block-rigid time-frequency pulse entanglement method, the envelopes 31 for four different f2 excitation pulses with frequencies of 1.5, 4, 2, and 3 kHz are shown. These envelopes are used to switch the excitation tones on and off; the pulse shape in this representation is only shown by cosine ramps without a steady state. Fig. 3 It shows, so to speak, a snapshot of a measurement block for a panel.

[0169] In Fig. 4The distribution of the four pulse pairs grouped in panel A across the four slots of the block is shown.

[0170] In Fig. 3b For the block-rigid time-frequency-pulse entanglement method, the envelopes 31 for four different f2 excitation pulses with frequencies 1.5; 4; 2; 3 kHz and the envelopes 32 for the four different f1 excitation pulses corresponding in the pulse pairs with frequencies 1.25; 3.33; 1.67; 2.5 kHz are shown. Fig. 3bThis represents a block with four pulse pairs, presented in four slots (Slot 1 to Slot 4), each 40 ms long. As can be seen, the excitation frequencies f1 of the pulse pairs are switched on at the beginning of each slot and switched off again at the end of each slot. Due to the slot length, or rather the switch-on duration of the f1 excitation pulses, these form plateaus, as can be seen from their envelopes. The four different f2 excitation frequencies, presented in four excitation pulses within the four slots, are only switched on after the f1 excitation pulses have been switched on. Furthermore, the f1 excitation pulses are only switched off after the f2 excitation pulses have been switched off and have already decayed. The switch-on and switch-off processes of the individual excitation pulses f1 and f2 are represented by cosine-shaped ramps.In contrast, the plateaus of the f1 pulses form a steady state, which lies between the cosine-shaped ramps of the switch-on and switch-off processes. The block time TB in the depicted block of four pulse pairs f1, f2 is 160 ms, which corresponds to the sum of the four slot times TS (4 x 40 ms). As described in paragraphs

[0023] to

[0024] and

[0050] , the presentation of short f2 pulses (or the presentation of short pulses for at least one of the two tones of the pulse pair) allows the separation of two signal components in the time signal of the distortion product, which, with the otherwise usual continuous presentation, can lead to distorted measurement results due to interference.

[0171] Panel A of the four f2 pulse pairs is presented in a 160 ms block, with each f2 pulse pair occupying a 40 ms slot. This block is repeated until a pre-selected signal-to-noise ratio (SNR) is achieved for all four measured DPOAEs.

[0172] This choice of frequencies and their arrangement are expedient for two reasons: A) Masking and interference

[0173] Frequencies in half-octave steps, i.e., for slots i = 1; 3 and i = 2; 4, are arranged with maximum spacing within the block, since the frequency ratio f2 / fdp is approximately 1.5. This leads to two types of interference dominating in principle – in the present example for i = 1;3: I) The presentation of the f2(i=3) pulse can mask the fdp(i=1) response on the cochlea if it has not yet decayed sufficiently. However, this only applies to the second source, which is often not the primary focus of the diagnostic procedure. II) The presentation of the f2(i=3) pulse can mask the f2(i=1) pulse on the cochlea because they are only about a minor third apart. Here, the primary contribution to the DPOAE, which is usually more important diagnostically, is also affected. Therefore, the time interval of T / 2 = 80 ms ensures that the interfering or masking signal components have typically decayed to an acceptable level. B) Time required

[0174] In this block-based method, averaging must continue until the frequency with the worst signal-to-noise ratio (i.e., the highest artifact rate) reaches the required minimum signal-to-noise ratio. In this example, this will be the measurement at f2 = 1.5 kHz. However, its noise background is typically no more than 30% worse than at frequencies f2 = 2, 3, and 4 kHz, so the time loss remains within acceptable limits.

[0175] Fig. 5 The diagram shows the distribution of seven excitation frequencies f2 across two panels A and B, each with four slots of 40 ms. Panel A is processed first; that is, for each f2, the mean sound pressure level Ldp of the corresponding DPOAE is determined with the desired SNR.

[0176] Panel B is then processed in the same way. The block containing Panel B also has a block duration of 160 ms, but this could be reduced to 120 ms because the frequency intervals between f2 and fdp are sufficiently large. 2.3.1.2 Block-flexible time-frequency pulse entanglement method with fixed pulse arrangement

[0177] If a frequency range is to be covered in which the noise background or the artifact frequency varies significantly, or if the procedure is to run optimally for each individual patient, then a large variability regarding the required averaging time for the respective frequency f2 must be assumed.

[0178] One solution to this problem is a block-flexible method for arranging pulse pairs. The frequencies used are, for example, f2 = 1; 1.5; 2; 3; 4; 6; 8 kHz. Under normal circumstances, the measurement at f2 = 1 kHz will take four times as long as at f2 = 2; 3; 4 kHz, because the noise background is twice as high, at least if the same noise background is to be achieved.

[0179] There are essentially two strategies for responding to this situation: First, one can accept the higher noise background. The consequence is that the estimation accuracy is not as high at low frequencies, since this noise is taken into account when determining the point closest to the threshold in the growth function during threshold approach, resulting in a higher extrapolation error. Second, one can try to allocate more measurement time to the frequency with the increased noise background.

[0180] The frequencies are therefore divided into three panels A, B and C, each presented in a block with 4 slots: in slot 1 the pulse pair at f2 = 1kHz is presented permanently, i.e. for all three panels; in slot 3 the pulse pair at f2 = 1.5kHz is presented for 3 / 4 of the time, and the pulse pair at f2 = 2kHz for the remainder of the time; in slot 2 the pulse pair at f2 = 3kHz is presented for the first half of the measurement time, and the pulse pair at f2 = 8kHz for the second half; and in slot 4 the pulse pair at f2 = 6kHz is presented for the first half of the measurement time, and the pulse pair at f2 = 8kHz for the second half.

[0181] Fig. 6The diagram shows the distribution of the seven f2 pulse pairs across three panels: A, B, and C. Panels A, B, and C are processed sequentially. In panel A, the mean values ​​for Ldp at f2 = 3 kHz and f2 = 6 kHz are determined with sufficient SNR. In panel B, the measurement continues at f2 = 1 kHz, is completed at f2 = 1.5 kHz, and begins at f2 = 4 kHz and f2 = 8 kHz. The measurement time for panel A is longer than that for panel B. In panel C, the measurement is completed at f2 = 1 kHz, f2 = 4 kHz, and f2 = 8 kHz, and begins and ends at f2 = 2 kHz.

[0182] In such an arrangement, the different noise and artifact conditions for the various excitation frequencies are taken into account; it has been ensured that the presentation of f2(i + 1) a half octave higher than f2(i) never directly follows the latter, and in addition, in slots 2 and 4 the total measurement time will be divided asymmetrically among the three panels, so that the high frequencies, f2 = 6; 8 kHz, are allocated more averaging time when the noise background increases slightly again at these frequencies.

[0183] Here, the first excitation level is processed first, followed by the second excitation level in the same arrangement according to the threshold approximation method, etc.

[0184] This block-flexible method can also be suboptimal, especially if the DPOAEs differ significantly, indicating a substantial hearing impairment. A higher level of pulse entanglement can therefore only be achieved with a free, situation-adapted pulse arrangement. 2.3.1.3 Block-flexible time-frequency-pulse entanglement method with free pulse arrangement

[0185] Here, jobs are assigned to the measurement of growth functions using the threshold approximation method. If DPOAEs are measured for seven frequencies f2 in half-octave steps, then seven jobs are processed. The measurements are expediently performed with excitation levels L2 that are as similar as possible to reduce masking problems. Therefore, all jobs start with the level that will be furthest from the threshold for people with normal hearing.

[0186] These seven jobs are not necessarily started in parallel. The measurement begins with panel A. Panels B through D are not yet assigned; they are determined by the automated reassignment of a slot after computer unit 15 has determined that a corresponding DPOAE for a pulse pair measured in panel A has a sufficient SNR.

[0187] The Figs. 7 and 8 The process of reassignment is illustrated as an example. The jobs are according to Fig. 7 The jobs are distributed across four panels A, B, C, and D, each with up to four slots within a block. The current arrangement of jobs—that is, pulse pairs of different f2—in the four slots is called a panel and must conform to a set of criteria that, for example, take masking effects into account.

[0188] Estimated runtimes for the slots are entered into a matrix. At the start of the measurement (Panel A), the estimates are based on empirical data from the population dominant for the diagnostic task. During the presentation of the individual blocks, all jobs are monitored. As soon as a job reaches its termination criterion (sufficient SNR or reaching the maximum presentation time), the estimated processing times of the slots are recalculated, and the vacated slot is reassigned to a pulse pair that still needs to be processed.

[0189] In this example, the initial setup consists of the same arrangement as described above in section 2.3.1.2, but is varied below. For the initial setup, it was assumed that measurements at f2 = 1 kHz would take four times as long as those at f2 ≥ 2 kHz due to the doubled noise level, and those at f2 = 1.5 kHz would take twice as long.

[0190] Then the chosen setup is nearly optimal; theoretically, time is wasted in panel C, as the measurement will end prematurely at 2 kHz. The crucial factor is the switch between panels B and C: here, the number of slots was reduced to two, as the necessary one-octave spacing is maintained at the two remaining frequencies, allowing twice as many pulses to be presented per unit of time compared to the 4-slot system. Alternatively, theoretically, the same measurement time could be achieved with a continuous 2-slot system.

[0191] If, after processing jobs 2 and 7 (f2 = 3 and 8 kHz respectively), it is determined that they are running significantly ahead of schedule or that the jobs in the remaining slots are running with a delay, the system checks whether an unprocessed job can be taken over. In this case, that could be job 3 (f2 = 2 kHz). However, if it is processed, only job 1 will remain, so no time will likely be saved. But as soon as the jobs in slot 4 are also completed prematurely, the system can switch to the 2-slot system earlier, and after processing job 3 (f2 = 2 kHz), the measurement can be ended with a single measurement at 1 kHz.

[0192] While in Fig. 7 For the block-flexible time-frequency-pulse entanglement method with free pulse pair arrangement, the distribution of a total of seven f2 pulse pairs across four panels and four blocks with a variable number of slots was shown under the simplified assumption of uniform measurement times for selected f2 pulse pairs. Fig. 8a variation in which no uniform measurement times are required.

[0193] As soon as a slot becomes available because the mean Ldp with sufficient SNR has been determined for the f2 previously measured there, it is filled with a new f2, resulting in a new panel. This can shorten the measurement time for individual panels. This is particularly relevant for hearing-impaired patients because the new method does not require any a priori assumptions about the expected SNR or the necessary measurement time of the respective DPOAE.

Claims

1. Method for the examination of the hearing ability for at least one ear of a mammal, in which growth curves are determined on the basis of the measurement of DPOAE's evoked by pairs of excitation signals (f1, f2) for various excitation frequencies f2, including the steps: the ear is presented with first excitation signals with a first excitation frequency f1 and a first noise level L1 and second excitation signals with a second excitation frequency f2 and a second noise level L2 at the same time or at temporal offset; and pulse pairs with a first pulse of the first excitation signal and a second pulse of the second excitation signal are presented in the ear, and the DPOAE's evoked thereby are captured and evaluated; characterized in that a block which has at least two different pulse pairs with different second excitation frequencies f2, is repeated several times during a measuring period, wherein the duration of the first and the second pulse in a pulse pair is 2 to 20 ms, and wherein each block of pulse pairs is presented during a block time TB, which is selected such that there is a time interval of 30 to 100 ms between the start of a first pulse pair and a subsequent pulse pair with the same excitation frequency f2, wherein, within a block, the start of a pulse pair follows the start of the immediately preceding pulse pair in the block by a time interval T, where T corresponds at least to the duration of the preceding pulse.

2. Method according to claim 1, characterized in that, in one block, the second excitation frequencies f2 of two immediately consecutive pulse pairs are at least one octave apart.

3. Method according to any one of claims 1 to 2, characterized in that the measured noise levels of the DPOAE's for pulse pairs of the same second excitation frequencies f2 are averaged during the measuring period.

4. Method according to any one of claims 1 to 3, characterized in that the or every block of pulse pairs during a block time is presented that is chosen in order to have a time interval of at least 70 ms, between the start of a first and a consecutive pulse pair with the same excitation frequency f2.

5. Method according to any one of claims 1 to 4, characterized in that, at the beginning of the measurements, a check occurs to determine whether the frequency fdp of one of the DPOAE's interferes with a spontaneous emission (SOAE).

6. Method according to any one of claims 1 to 5, characterized in that, within one block, the second noise levels L2 of the pulse pairs have a level difference from one another that is smaller than 15 dB.

7. Method according to any one of claims 1 to 6, characterized in that the sequence of the pulse pairs and the time interval between two immediately consecutive pulse pairs in a block remains constant.

8. Method according to any one of claims 1 to 6, characterized in that, once a desired signal-to-noise ratio for an excitation frequency f2 has been reached, the further averagings planned for this excitation frequency f2 are skipped.

9. Method according to any one of claims 1 to 8, characterized in that the noise levels of the DPOAE's for all the second excitation frequencies f2 contained in the block are measured and averaged for a second noise level L2 respectively allocated to the excitation frequency f2, and the measurements are conducted at least once for new noise levels L2.

10. Method according to claim 9, characterized in that, for each excitation frequency f2, a growth curve of measured values of the noise levels of the DPOAE's is determined for various noise levels L2, and the respective threshold values are then determined from said growth curves.

11. Method according to claim 10, characterized in that the growth curve is determined from at least three values for the noise level Ldp of the DPOAE determined at three different noise levels L2, but the same excitation frequency f2, wherein the at least three noise levels L2 include an upper noise level L2 that is used to determine a lower noise level L2, wherein the third, middle noise level is determined by means of the upper and the lower noise levels.

12. Device configured for executing the method according to any one of claims 1 to 11, comprising at least one ear probe (11) to be placed on / in the ear with one or two miniature loudspeakers (16) and one receiver (17), wherein the one or every miniature loudspeaker (16) is designed for the presentation of a first excitation signal with a first excitation frequency f1 and a first noise level L1 and / or a second excitation signal with a second excitation frequency f2 and a second noise level L2, and wherein the receiver (17) is designed for the capture and forwarding of a DPOAE evoked by the first and second excitation signal, wherein, furthermore, a computer unit (15) is provided that is programmed and configured to present a set of at least two pulse pairs with different second excitation frequencies f2 in a block, which is run repeatedly several times for one measuring period.

13. Hearing aid with a device according to claim 12.

14. Use of the method according to any one of claims 1 to 11 and / or the device according to claim 12 in a method according to any one of claims 1 to 11 for adjusting a hearing aid.

Images