Audiological Bulletin no. 35

Transcription

1 Auiological Bulletin no. 35 Ensuring the correct in-situ gain News from Auiological Research an Communication / 05-07

2 Introuction Hearing ais are commonly fitte accoring to ata base on a stanar ault ear. This neglects the fact that the acoustic conitions aroun an iniviual s outer ear are rarely stanar. The ifference between stanar an iniviual outer ears can be seen in the RECD. If we want to be sure to present the correct soun pressure at the earrum, this ifference between average ata an real-life iniviual ata must be measure an correcte for in the hearing ai fitting. To o this, Wiex measures the iniviual hearing threshol with the hearing ai in the ear, the Sensogram, an with the introuction of the Assessment of In-Situ Acoustics concept (AISA) we come one step closer to perfection in fitting precision. Here is where also the iniviual RECD ata comes into the picture. Iniviual RECD s compare to average ata Many factors can make the iniviual RECD iffer from the average. The most prominent effect is the effect, which can have a large impact in the low frequency area as shown in the figure below. RECD [B] = 0 mm = 1 mm = 2 mm Ø Ø Ø = 3 mm Ø k k Frequency [Hz] Other factors coul be the imensions of the tubing connecting the hearing ai with the earmoul or the imensions of the ear canal itself. How oes RECD ata influence the hearing ai fitting? As RECD values have an influence on the amplification, a change in RECD values means a change in the soun perssure level when the soun is prouce by the hearing ai. This change will have an effect in all the circumstances where the hearing ai is use as a soun generator. The hearing ai acts as a soun generator in two ways: - when measuring the in-situ hearing threshol where the hearing ai prouces the test tones - when the hearing ai is in normal use, taking in soun from the environment an amplifiying it

3 Even more precise threshol measurement When the RECD is ifferent from what the hearing ai expects, the soun pressure level calculate by the fitting software when making the Sensogram measurement oes not correspon 100% to what is presente in the ear canal. Let us have a look at an example: A hearing ai user has an ear canal that iffers from the average ear: His RECD value at 250Hz is 20B lower than the efault average RECD value in the hearing ai ue to a in the earmoul. That means that the hearing ai elivers a soun pressure level that is 20B lower than what is was originally set to provie. The user s hearing loss tells us that he starts hearing a soun when the soun pressure level at the earrum is 50B HL. But as the hearing ai elivers a soun pressure level that is 20B lower ue to the, the Sensogram measurement is artificially raise to 70B HL to prouce this 50B HL real soun pressure level at this iniviual s earrum, which is the value at which he starts hearing a soun. To match the nees of the real iniviual ear an to measure the correct hearing threshol, this eviation from the average has to be known an correcte for. We can obtain this knowlege from either a effect measurement, as this low frequency eviation normally woul be explaine by a partly open earmoul, or by making an iniviual RECD measurement. When we know the egree of eviation from the average, it is easy to correct the measurement. In this example we measure the user s hearing threshol to be 70B HL. We have now estimate the effect to be 20B an therefore know that the soun pressure at the earrum is 20B lower than the efault hearing ai setting. We can then fin the correct threshol by eucting the known eviation from the measure threshol, an the real threshol will be: 70B HL 20B = 50B HL. Measurement an gain setting using Sensogram an correcting both gain an threshol. What the hearing ai expects How it is in the real worl RECD 0 B -20 B Threshol measurement 70 B HL 50 B HL Gain base on a 50B HL threshol an half gain rule (simplifie for the example) 25 B 5 B Correcte gain base on RECD ifference 45 B 25 B The correctly measure threshol gives the best starting point for calculating the neee target gain for all input levels.

4 Correct soun pressure level at all times The target gain is calculate from the measure threshol using the fitting rationale for the hearing ai. But if the hearing ai elivers less soun pressure ue to a than the manufacturer has esigne it to o, the gain setting in the hearing ai will also be too low. In the example we know that the presente soun pressure is 20B lower than planne at 250Hz. To get the correct soun pressure level, we therefore have to a an extra 20B gain at that frequency. To make the calculation easy, we use a half gain fitting rule in this bulletin, which means that for a hearing loss of 50B HL, the target gain nees to be half of the 50B, that is: 25B. The lowers the soun pressure level in the ear canal by 20B so the real gain with no correction will be 25B 20B = 5B. That is 20B too low compare to the target gain, so we set up the hearing ai to provie 25B + 20B = 45B to obtain the real gain of 25B which was neee originally. Why oes the Sensogram not take this effect into account? When using an in-situ threshol measurement like the Sensogram, a ifference in the RECD value concerns the threshol measurement as well as the actual gain setting. The example from earlier on using a Sensogram instea of an auiogram but without any corrections: Measurement an gain setting using Sensogram without corrections. What the hearing ai expects How it is in the real worl RECD 0 B -20 B Threshol measurement 70 B HL 50 B HL Gain base on a 70B HL threshol an half gain rule 35 B 15 B Correcte gain base on RECD ifference not correcte We see that the hearing threshol measure is 20B too high. This means that we will set the hearing ai to eliver a higher gain than what we actually woul want for the real 50 B HL hearing loss, which is not known at this time. But as the same 20 B ifference cause by the applies to the gain setting (35 B), the increase gain will again be lowere by the effect (35 B 20 B = 15 B). The result is, as shown, a real gain that is 10B lower than the target gain of 25 B.

5 If the fitting was one base on threshol measurement ata from a well calibrate auiometer, the measure threshol showe 50B HL an we chose to use that as basis for the fitting, the result woul be a real gain of 5 B instea of the target gain of 25 B, as shown in the table below. Measurement an gain setting using auiogram with no corrections. What the hearing ai expects How it is in the real worl RECD 0 B -20 B Threshol measurement not measure 50 B HL Gain base on a 50B HL threshol an half gain rule 25 B 5 B Correcte gain base on RECD ifference not correcte In this example the auiometer calibration fits the user perfectly, so the presente soun pressure in the threshol measurement is exactly as expecte. This will almost never be the case in real life even though it shoul be close. When using an in-situ auiogram measurement an estimating the error either using a effect estimate or making an RECD measurement, the auiometry measurement will always be calibrate exactly to the iniviual user. One step further: looking at compression To explain the principle, we have use the half-gain rule for the example in this bulletin. In non-linear hearing ais, the hearing loss compensation will be tailore accoring to the input level. For soft input levels, the gain will be high an for lou input levels the gain will be much lower. How will this type of correction affect a fitting with compression hearing ais? How oes this compare to just making an in-situ auiogram measurement an get one with it? In the figure below, the real output level from the hearing ai is shown as a function of the input level when both the gain setting an the threshol measurement are correcte for the effect. The graph shows the gain set accoring to the half-gain fitting rule at a 50B HL input level, which in this example is the level of normal speech. For 0B HL input the target is a 100% hearing loss compensation so the output level will be 50B HL at 0B HL input level. For lou input levels, the target is efine by the UCL value. The target for lou levels is to play back soun with an input level equal to the normal-hearing person s UCL (about 97B HL) at an output level equal to the UCL of the hearing ai user. The hearing ai user s UCL is calculate from the user s measure hearing threshol.

6 output level [B HL] UCL 107B HL 60 target gain 25B 40 HTL 50B HL input level [B HL] By correcting both the hearing threshol an the gain setting, the hearing ai user gets the neee gain at all input levels. output level [B HL] Target: fully correcte gain an threshol Base on Sensogram: not correcte Base on auiogram: not correcte input level [B HL] The rawing to the left shows the resulting output for three scenarios. The blue line shows the result when a Sensogram is measure with no knowlege of the effect. At the threshol the gain will correspon perfectly to the target. At normal speech input levels aroun 50B HL, the correction for the half gain rule will also be half of the effect. For lou inputs, the fitting rule will estimate that the user s UCL is 114B HL, an as a result of the effect, this will be presente as 94 B HL instea. As it shoul have been 107, this is an error of 13 B. The re line shows the result using an auiogram measurement alone. Using the correct threshol, but not making any corrections for the effect, gain will be 20B too low for all input levels compare to the green target line.

7 Summary In most cases there will be a ifference between the soun pressure level we want to present in the ear canal of a hearing ai user, an the soun pressure level that the hearing ai really presents. Using the iniviual earmoul an hearing ai to measure the hearing threshol brings this ifference into the equation an thereby brings the en result a lot closer to the wante target. The only way to get everything correct is to measure the ifference an correct for it both in the threshol measurement an the gain setting. This measurement can be one by making an iniviual RECD measurement for each user. As the primary factor controlling the RECD for low frequencies is the an leakages in the earmoul, estimating the effect for frequencies below 1kHz coul also be sufficient.