Comparing real-bone measurement microphone and skull simulator measurements for soft headband bone conduction devices in adults

The use of bone conduction devices has increased in recent years, offering effective solutions for patients with conductive or mixed hearing loss and limited usable hearing unilaterally. This growth necessitates more accurate verification tools to ensure optimal patient outcomes. The real-bone measurement microphone is an innovative technology that enables in-situ verification of various bone conduction device types and is under development by Audioscan. This study presents measurements made using this technology on 10 adults wearing two commercially available soft headband bone conduction devices. Comparable measurements were conducted with the Verifit Skull Simulator, a commercially available technology for verifying abutment-based bone conduction devices. Force levels recorded with both technologies are compared. Results include real-head-to-dial difference and dB force level measurements across various scenarios. Average real-head-to-dial difference values measured using two bone conduction devices differ at most frequencies, while following the same overall trend as frequency changes. For the dB force level results obtained from the real-bone measurement microphone and Verifit Skull Simulator, the Pearson correlation between these two devices is consistently greater than 0.87, and the root mean square error varies between 6 to 12 dB across different test scenarios. The potential sources of differences between the two devices are analyzed and discussed in this paper. Findings provide insights into the performance characteristics of both technologies, demonstrating that the real-bone measurement microphone is a promising tool for expanding clinical verification capabilities across a wide range of bone conduction devices.

The implementation of bone conduction devices (BCDs) has seen consistent growth in recent years (Reinfeldt et al., 2015, 2022; Persson et al., 2020). These devices have proven effective in enhancing audibility and improving patient outcomes for individuals with conductive or mixed hearing loss, as well as those with limited usable hearing unilaterally (Reinfeldt et al., 2022). Unlike traditional air conduction devices, which rely on the middle ear to transmit sounds to the cochlea, BCDs bypass the middle ear by vibrating the skull and delivering vibrations directly to the cochlea (Reinfeldt et al., 2015). As the adoption of BCDs continues to grow, the need for standardized and precise verification methods for these devices becomes increasingly important.

Verification of BCDs is inherently more complex than air conduction devices due to fundamental differences in the sound delivery mechanisms and complexities of measuring device-generated vibrations. These complexities increase further when different BCD connection types, such as transcutaneous and percutaneous systems (Reinfeldt et al., 2022), are factored in.

Various approaches have been employed to verify the performance of BCDs, but each has notable limitations. One of these approaches is using aided sound field thresholds, which has its own advantages when it comes to the validation stage in the fitting process; however, it is known to have limited frequency resolution and poor test-retest reliability (Hawkins, 2004; Persson et al., 2023). Moreover, this approach provides only limited information on key aspects of BCD performance, such as their response to speech (or similar) inputs, input/output characteristics, and dynamic range (Hodgetts et al., 2018). Nasal sound pressure measurement has also been proposed in the literature for verifying certain types of BCDs (Reinfeldt et al., 2019), but this method is technically complex and impractical for routine clinical use (Persson et al., 2023).

One systematic objective approach to verifying BCDs that addresses several of the aforementioned limitations is to use skull simulators, which can measure the force level output of the hearing devices on an abutment (El-Naji et al., 2024). The Verifit ® Skull Simulator (Audioscan, Dorchester, ON, Canada) is a commercially available skull simulator for use in clinical settings. Although skull simulators demonstrate high test-retest reliability, they do not support all BCD connection types, which limits their clinical applicability for certain patients. Moreover, skull simulators cannot fully account for the complex interindividual variability among patients, which can play a role in the BCD verification process. This variability can be caused by factors such as variations in head size, bone thickness and density, and skin properties.

Given current limitations in the verification of BCDs, a system that allows verification while the device is worn by the patient has the potential to offer significant clinical benefits. This would make it possible to perform real-bone measurement (RBM) for BCDs, an approach equivalent to real-ear measurement (REM) used for air conduction devices, which can provide similar advantages to REM. RBM would account for interindividual variabilities among patients as it can be performed while the BCD device is being worn. This system would also enable measurements for a wide range of BCD technologies, regardless of connection type. At the time of writing this manuscript, the RBM microphone is under development by Audioscan and is expected to be available for clinical use in the near future. The RBM microphone represents the first device of its kind designed for clinical verification of BCDs using on-head measurements.

A few studies have explored similar concepts to the RBM microphone currently under development by Audioscan. For instance, Hodgetts et al. investigated the use of a surface microphone for verifying BCDs (Hodgetts et al., 2018). The study compared audibility measurements obtained with a surface microphone and a skull simulator in percutaneous BCD users. Their results suggest that, for percutaneous BCDs, the surface microphone can yield audibility values comparable to those obtained with a skull simulator. In another set of studies, Persson et al. (2023, 2025) also investigated a similar concept, which they referred to as a skin microphone, for the objective verification of BCDs. They conducted tests on both percutaneous and transcutaneous BCDs to measure audibility using a skin microphone, skull simulator and artificial mastoid. Based on their findings, they concluded that a skin microphone may be a promising tool for objective measurements of BCDs.

The goal of the current study is to present the results of measurements obtained using the RBM microphone in an adult population wearing soft headband BCDs, and to compare these results with those obtained using the Verifit Skull Simulator. This comparison is expected to provide valuable insights into the characteristics and clinical implications of each verification procedure.

Hearing devices and test stimuli

Two commercially available BCDs that support soft headband connection were selected from two manufacturers for the experiments in this study. These devices are referred to as BCD#1 and BCD#2 hereafter. Both BCDs were set to omnidirectional mode, and all noise reduction features were disabled. The built-in feedback reduction features of the devices remained active to reduce the likelihood of feedback during measurements. In-situ thresholds were entered into each manufacturer's fitting software at 20 dB dial level (DL) for audiometric frequencies between 250 Hz and 8 kHz, and the connection type was configured as soft headband. The output / gain characteristics of both BCDs were programmed to the first fit desired sensation level (DSL) BCD targets (Hodgetts and Scollie, 2017) provided by the manufacturer's fitting software, and no gain adjustments were made unless stated otherwise.

All tests were performed using a Verifit ® VF-2 and RBM microphone prototype compatible software. Speech tests were conducted at input levels of 55, 65, and 75 dB sound pressure level (SPL) using the long-term average speech spectrum (LTASS) of the international speech test signal (ISTS) (Holube et al., 2010). The LTASS represents a 1/3-octave spectrum obtained by averaging the acoustic energy of speech over an interval of at least 10 s to smooth out short-term fluctuations (Audioscan, 2025). Maximum power output (MPO) tests, using a tone burst stimuli, were performed in addition to speech tests to measure the maximum output of the BCDs in response to a steady narrow band input (Audioscan, 2025). In addition, the noise floor was also measured to assess the internal noise of the Verifit VF-2 when used with both the RBM microphone and the Verifit Skull Simulator. The noise floor measurement was done with both the BCD and Verifit VF-2 speaker muted.

RBM microphone measurements

A prototype version of the RBM microphone was used for measurements. The RBM microphone consists of two microphones: a measurement microphone positioned near and facing the patient's forehead, and an outward-facing reference microphone used for sound field measurement. The measurement microphone measures changes in pressure in a small volume created by vibrations of the patient's head when stimulated with a BCD. Additionally, the RBM microphone features a headband designed to secure the device on the head.

Each BCD was positioned on the participants' left mastoid with a soft headband, using typical clinical fitting procedures for band tension and positioning, with the tension adjusted so that the band was not too tight to cause discomfort, and not too loose to prevent device vibrations from being transmitted to the head. Subjects were positioned approximately 60 cm from the Verifit VF-2′s loudspeaker at 0 degrees azimuth, following the typical clinical setup used for REM with air conduction devices. For all tests (real-head-to-dial-difference (RHDD), speech, MPO, and noise floor), the RBM microphone was positioned at the center of the forehead as illustrated in Figure 1 and was connected to Verifit VF-2, with the band tension configured as with the soft headband device. The RBM microphone tests were performed in a simulated clinic room at the Audioscan facility.

For each BCD, all test scenarios (i.e., various stimulus types and levels) on each participant were repeated three times. After each repetition, the participant was asked to remove the RBM microphone from their head and place it back in the original position. For each participant and each test scenario, the result used in the analysis is the average of these three tests. This test repetition and averaging was done to increase reliability and stability of measurement results in the analysis.

All steps necessary for a complete RBM microphone measurement in the software were followed each time the RBM microphone was placed on a participant's head. This includes running a sound field leakage transfer function measurement, followed by an RHDD measurement, before conducting aided speech and MPO verification tests in Speechmap, a hearing instrument fitting environment in Audioscan devices. The sound field leakage transfer function measurement is an RBM microphone related feature, introduced by Audioscan. It is performed with the BCD muted and is designed to create a model of the extraneous sound leakage path into the RBM microphone from the presenting loudspeaker. This leakage path model is then used during subsequent speech tests to predict the actual sound field leakage and reduce its impact on the measurement, thereby improving accuracy. It should be noted that in the software prototype version used for the tests in this study, this leakage cancellation filter was applied to speech tests, but not to MPO tests.

The RHDD measurement represents the difference between the measured SPL at the RBM microphone and the in-situ DL used in the fitting software across frequencies for the BCD. The RHDD measurement was performed using the corresponding feature in the RBM microphone software, by playing tones at known dial levels at audiometric frequencies from the hearing aid. Any dial level that leads to a successful RHDD measurement at a particular frequency can be used and there is no need to use a fixed or specific dial level value. The measured RHDD is interpolated logarithmically to estimate values at other frequencies which are subsequently used in Speechmap.

The on-head force level in Speechmap was calculated using the measured SPL and RHDD, applying the transformation presented in Equation 1,

where DFL and RHCD stand for device force level and real-head-to-coupler difference, respectively (Hodgetts et al., 2018; Hodgetts and Scollie, 2017; Gascon et al., 2022). Both DFL and RHCD are properties of the BCD under test. In addition to the transformation described in Equation 1, a microphone location effect (MLE) was applied to the RBM microphone measurements to account for the acoustic difference between the position of the reference microphone on the RBM microphone (i.e., forehead) and the BCD microphone location (i.e., mastoid). The MLE was necessary because sound field equalization for speech and MPO tests for RBM microphone measurements was performed at the reference microphone on the RBM microphone rather than at the BCD position. The applied MLE was previously obtained on a manikin head in a standard clinical room without specialized acoustic treatment.

One other measurement note is that the MPO stimulus input level used during RBM measurements is 85 dB SPL given sound field speaker characteristics, whereas the MPO stimulus input level for Verifit Skull Simulator tests is 90 dB SPL inside the test box (Audioscan, 2025). To account for this input signal level difference, a 5 dB correction (increase) was applied to all MPO test results from the RBM microphone.

Participants

Data collection was performed on 10 adults (seven males and three females) recruited from among Audioscan employees. The mean and standard deviation of participants' ages were 43.1 and 11.62 years, respectively, with a range from 31 to 60.

Skull simulator measurements

The same BCD programs, stimulus types, and levels described in Section 2.1 were used for the measurements with the Verifit Skull Simulator. Each skull simulator measurement was repeated three times, and the reported results represent the average of these repetitions. As shown in Figure 1, the BCD under test was coupled to the abutment of the Verifit Skull Simulator that was connected to the Verifit VF-2 test box using the typical clinical approach outlined in the Verifit VF-2's users guide for measurements (Audioscan, 2025). In brief, the skull simulator was positioned with the abutment facing the left side of the box, and the reference microphone was placed approximately 1–3 mm from the BCD's front microphone opening. The test box measurements were conducted in the Speechmap environment. The predicted on-head force level in Speechmap was calculated using the measured force level from the skull simulator and the RHCD, as presented in Equation 2.

RHDD measurements

Figure 2 presents the average and standard deviation of RHDD values measured across the entire test population. The results indicate that the RHDD values measured using BCD#1 and BCD#2 differ across frequencies, with BCD#2 producing higher average RHDD values than BCD#1 except at 250 Hz and 1.5 kHz, where BCD#1 gives slightly higher average RHDD values. The most significant differences in average values occur at 750 Hz (approximately 10 dB average difference) and 8 kHz (approximately 22 dB average difference). Despite these differences, the average RHDD values for both devices follow a similar trend: values are higher at lower frequencies (ranging between 50 and 60 dB) and gradually decrease as frequency increases. Notably, BCD#1 shows a peak in the average RHDD values at 1.5 kHz. Across all audiometric frequencies and for both BCDs, the standard deviation of the RHDD remains below 6.5 dB for frequencies up to 4 kHz. The Wilcoxon signed-rank test with a significance level of 0.05 led to rejection of the null hypothesis that the distribution of differences in RHDD between BCD#1 and BCD#2 is symmetric about zero, except at 250 Hz and 3 kHz. This shows that the difference between the two devices is statistically significant at all other frequencies.

Speechmap measurements

Figure 3 presents the dB force level (FL) results averaged across the entire test population, obtained using the RBM microphone for all stimulus types and levels described in Section 2.1. It also shows the corresponding dB FL values measured under the same test scenarios using the Verifit Skull Simulator, along with their associated noise floor measurements. Additionally, the figure also illustrates the differences in dB FL values between the RBM microphone and the Verifit Skull Simulator for each test scenario.

The results in Figure 3 show that for all stimulus types and both BCDs, the dB FL values measured using RBM microphone and Verifit Skull Simulator follow a similar trend as frequency changes. Starting from 250 Hz, the dB FL values consistently increase with increasing frequency, reaching a peak around 900 Hz. Beyond this peak, the dB FL results gradually decrease with further increases in frequency. The Pearson correlation values were calculated over a frequency range of 250 Hz to 5 kHz, between the average RBM microphone dB FL measurements and Verifit Skull Simulator measurements. The calculated Pearson correlation values for ISTS 55 dB SPL, ISTS 65 dB SPL, ISTS 75 dB SPL and MPO were 0.91, 0.92, 0.95, and 0.91, respectively, for BCD#1, and were 0.96, 0.98, 0.97, and 0.87, respectively, for BCD#2. The Pearson correlation values were always greater than 0.87, confirming a strong correlation between the results as frequency changes.

The difference curves presented in Figure 3 show that for most test scenarios, there is a peak in the difference between average RBM microphone and Verifit Skull Simulator measurements around 250 Hz, where the difference ranges from approximately 10 to 20 dB. Additionally, across most test scenarios, another peak in the difference value is observed around 1,400 Hz. The root mean square error (RMSE) between RBM microphone and Verifit Skull Simulator are as follows: for BCD#1, for ISTS 55 dB SPL, ISTS 65 dB SPL, ISTS 75 dB SPL and MPO in the frequency range from 250 Hz to 5 kHz are 7.86, 8.10, 8.71, and 11.97 dB, respectively; for BCD#2, the corresponding RMSE values are 7.23, 6.04, 5.98, and 7.02 dB, respectively.

The Wilcoxon signed-rank test was used to assess whether the difference between the RBM microphone measurements (on the test population) and the Verifit Skull Simulator measurements were statistically different from zero across each test scenario and each audiometric frequency up to 4 kHz. A significance level of 0.05 was chosen and adjusted based on the Bonferroni method to minimize the type-I error. The null hypothesis that the difference equals zero was rejected for all conditions except the following: ISTS 55 dB SPL at 3 kHz for BCD#1; ISTS 65 and 75 dB SPL at both 3 and 4 kHz for BCD#1; MPO at 1 kHz for BCD#1; ISTS 55, 65, and 75 dB SPL at 1 kHz for BCD#2; ISTS 75 dB SPL at both 3 and 4 kHz for BCD#2; and MPO from 1 to 4 kHz for BCD#2.

In addition to comparisons between RBM microphone and Verifit Skull Simulator results discussed above, an analysis was performed to evaluate the test-retest reliability of RBM microphone measurements. As stated in Section 2.2, for each test scenario, RBM microphone measurements were repeated three times for each participant. For one of the scenarios, speech at 65 dB SPL for BCD#2, we calculated the standard deviation of these three tests for each individual. We then computed the average of these standard deviations across all 10 participants. The calculated average values (with ranges in brackets) at frequencies of 250, 500, 750, 1,000, 1,500, 2,000, 3,000, and 4,000 Hz are 3.84 (0.66–7.51), 1.76 (0.61–3.85), 0.90 (0.31–3.06), 1.08 (0.10–2.66), 1.01 (0.25–2.39), 0.99 (0.50–3.02), 0.82 (0.06–2.76), and 1.59 (0.15–6.03) dB FL, respectively.

This study presents RHDD and dB FL measurement results obtained using the RBM microphone on a population of adults and compares these results with those obtained from the Verifit Skull Simulator. As explained in Section 3, the trends of dB FL values obtained using RBM microphone and Verifit Skull Simulator are similar as frequency changes (Pearson correlation of greater than 0.87). This can be discerned from the results in Figure 3 and from the calculated correlation between RBM microphone and Verifit Skull Simulator across frequencies. That said, there are some differences in the exact dB FL values measured by these two technologies at various frequencies. The results presented in Section 3.2 showed an RMSE of 6 dB or more for both BCDs across various test scenarios. These discrepancies may arise from multiple sources. First, although the RBM microphone and the Verifit Skull Simulator are used to verify similar devices, the nature of their measurements differs. The interindividual variability and differences in participants' head anatomy are inherent to the RBM microphone measurements. These factors cannot be accounted for in the Verifit Skull Simulator measurements. Consequently, part of the observed differences between the RBM microphone and Verifit Skull Simulator results may stem from intrinsic differences in their measurement methods. In addition to these intrinsic differences in measurement procedures, in the next paragraphs we describe other non-intrinsic factors that may contribute to the observed differences across various frequencies.

As shown in Figure 3, the effects of external sound leakage are more pronounced in the MPO tests (in the low-frequency region). This may be because the sound field leakage cancellation filter described in Section 2.2 was applied only to the speech signals and not to MPO tests in the RBM microphone software version used for this study. Apart from this, one factor that may contribute to the differences observed between the RBM microphone and the Verifit Skull Simulator results for both speech and MPO tests at low frequencies (approximately 250–500 Hz in Figure 3) is the very low gain values prescribed by the DSL BCD at these frequencies, particularly at 250 Hz. These low gain values result in low-amplitude vibrations generated by the BCD. Due to their magnitude, these vibrations might be susceptible to interference from external low-frequency sounds that leak into the RBM microphone, which can affect the measurement results. To investigate the effects of gain on the RBM microphone's low-frequency response, an additional set of measurements was conducted for one participant. In this set, the gains provided by the hearing aid software were manually adjusted at 500 Hz and below, either increased by 20 dB or set to the maximum allowed by the BCD software when a 20 dB increase exceeded the maximum allowable value. This adjustment was applied to both BCDs used in the current study. The same test stimuli and levels described in Section 2 were used with measurements taken using both the RBM microphone and Verifit Skull Simulator. Figure 4 presents the differences between the RBM microphone and the Verifit Skull Simulator dB FL results using both the original and low-frequency-modified gain settings. As these results show, across all test scenarios, increasing the low-frequency gain leads to a reduction in the low-frequency dB FL differences between the RBM microphone and Verifit Skull Simulator, supporting the justification provided here.

Since the RBM microphone tests were not conducted in an acoustically insulated environment, undesired effects from sound reflection may have been introduced. These reflections could have influenced the RBM microphone measurements and potentially affected the accuracy of our comparison between the RBM microphone and Verifit Skull Simulator results. These effects may be present for any of the test scenarios and at various frequencies. As also explained in Section 2.2, for the RBM microphone measurements, the sound field equalization was performed at the location of the RBM reference microphone (i.e., forehead) rather than at the location of the BCD microphone (i.e., mastoid). As such, a related factor that may also contribute to differences between RBM microphone and Verifit Skull Simulator results presented in Figure 3 is the MLE applied to the RBM microphone. As described in Section 2.2, the MLE applied to our measurements was obtained from a manikin head in a room with no specialized acoustic treatment. As such, differences observed in our findings between RBM microphone and Verifit Skull Simulator may reflect the differences between manikin head and individual subject head characteristics, along with the influence of room acoustics. These differences can affect all frequencies within the bandwidth of interest, including the peaks observed in the difference curves in Figure 3, particularly in the 1–2 kHz range.

For the MPO results, as mentioned in Section 2.2, a correction of 5 dB was manually applied to the RBM microphone results across frequencies to account for the difference in the MPO stimulus input level for the RBM microphone and test box (skull simulator) measurements. This manual correction may introduce additional errors into our analysis as the 5 dB increase in the RBM results may not precisely correspond to 5 dB due to compression characteristics of BCDs.

The results presented in Figure 3 show that the average noise floor values for the RBM microphone start to increase at frequencies above 2 kHz. As a result of this increase, the noise floor can interfere with the RBM microphone force level measurements at high frequencies. The frequency region over which the RBM noise floor can impact the measurement results depends on the input level, with a generally broader measurement bandwidth as input level increases. As the results show, the high-frequency differences between RBM microphone and Verifit Skull Simulator measurements are greater at softer input levels, where the RBM measurements are masked by the noise floor in the higher frequencies, in contrast to the Skull Simulator measurements.

One other factor that needs to be highlighted is that, although the sample size in this study was considered appropriate for the scope of this work, a limitation of the current study is the relatively small sample size, which may restrict the generalizability of some conclusions. Related to this, the test population in this study only included adult participants, whereas including other age groups would be helpful for future studies to compare and investigate results between adults and children. One other point worth noting is that the performance and characteristics of the RBM microphone with connection types other than soft headband, such as percutaneous or active transcutaneous systems, need further investigation.

Overall, the results obtained using the RBM microphone and the Verifit Skull Simulator exhibit similar trends across frequencies, particularly as shown in Figure 4. It should be noted that, as the RBM microphone is currently under development, the experiments conducted for this paper utilized a prototype version of RBM microphone hardware and software, as available at the time of writing. Therefore, the results presented in this paper may not fully reflect the performance of the final production version of the RBM microphone. Overall, the emerging data point to the RBM microphone as a promising path toward more comprehensive and configuration-inclusive BCD verification.