A variety of approaches have been taken over the years to attenuate, filter, or otherwise reduce the distracting effects of background noise for the hearing aid user. This has long been a priority for hearing aid manufacturers. This article summarizes the approach taken by Interton to address this issue, and presents data gathered in a field trial recently conducted at the University Hospital in Giessen, Germany. The field trial was based on the evaluation of Intertons Speech Management System (SMS), a system that is designed to increase listening comfort for the user when in the presence of background noise.
Figure 1. Speech Management System (SMS) functionality: The upper panel shows an unprocessed mixed speech and noise signal. The lower panel shows the same signal processed by SMS. The lower panel indicates that SMS decreases the noise energy when no speech (no modulation) is present. When speech is present, modulation is detected, and the energy of the signal remains constant.
SMS works by measuring the modulation of the input signal (Figure 1). When speech is present, modulation is high, meaning that there are frequent peaks and valleys in the waveform. Under these conditions, the system is not implemented. When noise is present, however, modulation is typically low and levels are typically steady. SMS interprets this reduced modulation as noise, and attenuates the signal for as long as this condition exists. When modulation increases, as when speech is reintroduced, the system restores appropriate gain.
In this way, the system is essentially a comfort feature intended to increase ease of listening in noisy situations. SMS is not specifically intended to improve the signal to noise ratio (SNR), but essentially does so when the noise spectrum is dominated by frequency regions that are not crucial to speech recognition. The system functions independently in each channel of the multichannel device, and the amount of attenuation provided in each channel is dispenser adjustable. While the acoustic characteristics of this system are well known, the potential benefit of various combinations of the system settings are largely unknown. The primary motivation for the following field trial was to determine this benefit.
The study was conducted at the Audiology Department of the University Hospital in Giessen, Germany. The test device was the Quantum Evo Plus, a first-generation multidigital hearing instrument. This hearing instrument consists of a 3-channel 7-band WDRC system, with SMS functioning in each channel. The device also includes AGCo, microphone noise reduction (expansion), and multiple memories, the latter serving as a convenient way to evaluate multiple test conditions.
The study was designed to answer the following questions:
- What combination of SMS settings will be preferred in a specific listening situation (namely, traffic noise)?
- Is there an indicator in the individual patient data that suggests a particular SMS setting for a particular listening situation?
- What is the test-retest reliability of a subjective SMS optimization procedure?
- What is the benefit of SMS in everyday life situations?
Figure 2. Audiometric data of hearing impaired subjects.
Subjects. Nineteen subjects (16 male, 3 female ages 49-76) with moderate to severe sensorineural hearing losses participated. The mean audiogram associated with test subjects is shown in Figure 2. All subjects were experienced hearing aid users, had experience with psychoacoustic listening tasks, and were paid for their participation. Testing was conducted over three test sessions, as described below:
Session 1. During the first test session, audiometric data (audiogram, UCLs) were acquired and the hearing aid was fit to each participant. In most cases, the fitting was binaural. The initial fitting method utilized Intertons proprietary fitting algorithm, Dynagraph, a non-linear fitting method intended for WDRC-based applications. For this phase of testing, SMS was not used.
After the initial fitting the hearing aids were fine-tuned based on further testing. First, the subject was presented with the Giessener Dialogue, a recording of a male and female talker made in quiet and in a background of noise babble. Fine-tuning adjustments were made in order to maximize sound quality for both types of environment (ie,. in a single memory). The subject was then given a structured walk through the hospital to verify that settings were reasonably appropriate in a variety of situations. Once completed, final settings were designated FT1. The three memories of the device were then programmed as FT1, FT1-3 dB and FT1-6 dB. At this point subjects were given a diary to record impressions of FT1 at each gain setting, as experienced in the real world. Subjects then began a three-week acclimatization period that was designed to familiarize subjects with the gain, frequency response, and nonlinear (ie, compression) aspects of their aids without introducing SMS.
Session 2. Following the acclimatization period, the preferred gain setting was determined and any necessary fine-tuning adjustments (based on the diary, a repeat of the Giessener dialogue, and another hospital walk-through) were made. These final settings were designated FT2.
Once FT2 was established, an SMS optimization procedure was conducted. While listening to FT2, a sample of traffic noise was presented in the soundfield, and the subject was asked to make a paired comparison of two sets of SMS settings. For each pair, the subject was asked if a difference could be heard and, if so, which provided better sound quality or comfort.
The SMS optimization procedure used the four possible SMS settings (Off, Low, Medium, and High, corresponding to 0 dB, 6 dB, 12 dB, and 18 dB of attenuation) in each of the three signal processing channels. The first comparison was between no SMS in any channel (baseline) and one of the variants. Note that each variant represents only one step in one channel compared to the baseline condition. If a difference/preference was noted, this would become the baseline, and another single step contrast to it would be evaluated. If a difference was noted that proved negative, this condition was removed and another introduced. In this manner, a wide variety of comparisons were made until the optimized response for traffic noise was achieved.
Once the optimization procedure was completed, the three memories of the device were programmed as follows: 1) FT2 with SMS deactivated, 2) FT2 with SMS optimized, 3) FT2 with SMS set to default settings for a background noise environment (Low in each channel). The different programs were randomly distributed among the three memories of the hearings aids so that only the tester was aware of the memory allocation.
Once programmed, subjects repeated the three-week acclimatization period and noted their impressions in a variety of listening environments. After the test period, the participants were asked to judge the performance of the devices in various listening situations and to specify the program that was most beneficial to them as their listening environments changed.
Session 3. After the second acclimation period, subjects were invited for a final session in which the SMS optimization procedure was repeated in order to measure the test-retest reliability of the method.
Figure 3. The absolute frequency of the chosen SMS-level in the Low frequency channel (LB), Mid-frequency channel (MB), and High-frequency channel (HB) for the 38 ears.
Optimization Setting. Figure 3 provides an overview of the optimal settings chosen by subjects (38 ears) for the listening situation Traffic Noise. These results are consistent with expectations: optimal settings were generally composed of SMS on (typically on Low) in the low and middle bands, and typically off in the high band. This pattern is consistent with the spectrum of the traffic noise signal, which was dominated by low and mid frequency energy.
If the two most-selected settings are noted in each channel, the following conclusions may be drawn:
- Low-frequency channel (0 Hz 750 Hz): The preferred SMS level lies between low and medium, representing an attenuation of between 6 dB-12 dB.
- Mid-frequency channel (750 Hz 3000 Hz): The preferred SMS level lies between off and low, representing an attenuation of between 0 dB-6 dB.
- High-frequency channel (3000 Hz 8000 Hz): The preferred SMS level lies between off and low, representing an attenuation of between 0 dB-6 dB.
Table 1. Frequency of the activation/deactivation of the SMS in different channels depending on the average hearing loss in the corresponding channel.
To further understand the nature of preferred SMS settings, results were analyzed with regard to severity and slope of the audiogram. These results are summarized in Table 1, which shows the frequency of SMS activation for milder (<55 dB HL) and more severe (>55 dB HL) hearing thresholds in the low, middle, and high frequencies. The question was simply whether or not SMS was activated in each channel (ie, but not which level of attenuation was chosen). Table 1 indicates that, for milder losses in the low and middle frequencies, SMS was generally activated in the low and middle channels once SMS settings were optimized. This was also true for more severe losses in the low frequencies, where each of the seven subjects in this category preferred SMS activated in the low band. In the high band, data were nearly evenly split between SMS on and SMS off, regardless of audiogram; however, once the SMS settings were optimized, the results generally indicate that SMS was usually turned off in the high band.
Table 2. Frequency of the activation/deactivation of the SMS in different channels depending on the average gain for loud input signals (G-80) in the corresponding channel.
Results were further analyzed with regard to the amount of gain for loud (80 dB SPL) sounds provided by the FT2 memory setting. It was thought that those subjects with loudness recruitment, who receive less gain for loud sounds as part of the Dynagraph fitting algorithm, might also prefer the use of the speech management system more than those who can tolerate higher gain. Table 2 summarizes the results of the activation/deactivation analysis using loud sound gain (G-80) as an indicator. The data shows that, when Dynagraph/FT2 called for 5 dB or less gain for loud sounds, SMS was nearly always in use following its optimization, but only in the low and middle channels. In the high frequency channel, the percentage of SMS activation decreased considerably.
Tables 1 and 2 suggest that both the audiogram and the G-80 parameter may be used to determine whether SMS should be activated for traffic noise (and most likely other types of noise with similar low-frequency weighted spectra). Results suggest that losses ranging from relatively mild to rather severe in the low and middle frequencies will benefit from SMS activation. More severe losses will result in lower gain requirements for high input levels, which in turn also suggest the use of the system. These results indicate that, when a low and mid frequency weighted sound is presented at relatively high levels to individuals such as these, the system will provide its intended benefit of improving listener comfort.
The criterion for SMS activation/deactivation in the high frequency channel is somewhat more difficult to establish. This can perhaps be explained by the spectral characteristics of the sound sample chosen. If a broad spectrum noise or one with a high frequency emphasis were to be used, further insights might be gained.
Figure 4. The difference, in number of attenuation steps, between the test and retest SMS optimization results (n=38 ears).
Test-Retest Reliability. In the final session, the SMS optimization procedure was repeated in order to measure test-retest reliability. The reason for repeating the optimization procedure was to determine whether the three-week acclimatization to the system would result in changes to preferred settings. Figure 4 shows the difference, in the number of attenuation steps (from low to medium, for example), between optimization Sessions 1 and 2. Results indicate that reliability was quite high in the low and the high frequency channel, with most subjects making no changes in preferred settings. In the mid frequency channel, preferred settings differed in 50% of the cases by about one step (6 dB). In a more detailed analysis it was observed that, of the subjects who changed their settings during optimization Session 2, 85% had FT2 settings that called for loud sound gain (G-80) of 5 dB. This further supports the conclusion that the parameter G-80 and the preferred system setting are related.
Figure 5. Relative frequency (38 ears) of preferred programs: 1) FT2; 2) FT2 and optimized SMS; 3) Standard Party program based on FT2 when qualifying the listening situations Quiet, Radio/TV, Small Groups, and Automobile. Not experienced means that this situation did not occur for a subject during the test period.
Benefit in Every-day Situations. As indicated previously, subjects did not know what was programmed into each memory of their hearing aids; they were simply asked to try each program in a variety of situations and to note which program seemed to be the most helpful. In Figures 5 and 6 each of the three conditionsFT2 alone, FT2 + optimized SMS, and FT2 with SMS set to low in each channel (the default settings for a background noise algorithm)were evaluated in a variety of listening situations. Figures 5-6 show the frequency of preference for a given condition in a specific environment. The figures also reflect those occasions when a particular environment was not experienced by a listener during the trial period.
Figure 5 shows that, in Quiet situations, FT2 alone was the preferred condition, a result that was expected. The same result occurred for the situation Radio/TV, which normally takes place in quiet surroundings. In Small Group settings the standard Party algorithm (SMS low in all channels) was used most frequently. In an automobile, however, FT2 plus optimized SMS was clearly preferred.
Figure 6. Relative frequency (38 ears) of preferred programs: 1) FT2; 2) FT2 and optimized SMS; 3) Standard Party program based on FT2 when qualifying the listening situations Large Group, Party, Sound Quality, and Overall Impression. Not experienced means that this situation did not occur for a subject during the test period.
Figure 6 shows that, if the group size increased (Large Groups) or if the listener was in a party-like situation (Party), the program with the optimized SMS setting was also preferred. This result is somewhat surprising, as the system had been optimized specifically for Traffic Noise, but it also shows that it functions effectively in a variety of environments with spectra similar to that of traffic. These results suggest that the system increases listening comfort in those environments dominated by low and mid frequency noise.
When asked about general Sound Quality, subjects indicated that maximum sound quality was achieved in quiet environments when FT2 alone was engaged. FT2 combined with the optimized system was also highly rated, which suggests that, when used in its intended (noisy) listening environments, sound quality should be quite acceptable when SMS is used. The Overall Impression of subject results indicate that FT2 and FT2 + optimized SMS account for the preferred condition by nearly 75% of subjects. In general, these results suggest the importance of individual fine tuning in terms of gain, frequency response, compression, and special features such as SMS.
The SMS feature of Intertons hearing instruments has been investigated in a Quantum EVO BTE with regard to a specific type of listening environment, namely Traffic Noise. After an acclimatization period in which subjects familiarized themselves with general characteristics of the fitting (without SMS), SMS settings in the three channels of the test device were introduced and fine-tuned by means of a paired comparison optimization procedure. Once optimized, settings were analyzed with regard to the audiogram and to the amount of prescribed gain for loud sounds. Subjects were then asked to evaluate their hearing instruments in normal daily situations.
Data analysis shows that the preferred SMS setting depends on two factors: the amount of hearing loss and the amount of prescribed gain for loud sounds (G-80) in each of three frequency regions, corresponding to three signal processing channels. These factors indicate that the system will be beneficial for a variety of hearing losses, but especially those that exceed 55 dB. It will also be most beneficial when the amount of prescribed gain for loud sounds is less than or equal to 5 dB. Based on 38 ears, the preferred SMS settings for traffic noise were found to be between Medium and Low in the low frequency channel, and between Low and Off in the mid and high frequency channels.
When evaluated in the field by subjects, the environments in which the optimized SMS settings seemed to be particularly advantageous were for Automobile, Large Groups, and Partiesacoustic situations that share a similar spectrum. This suggests that the benefits of system optimization generalize across listening environments with similar acoustic properties. It should also be noted that the system optimization condition was preferred in these environments over default settings intended to be used in background noise.
The results of this study are based on a procedure intended to optimize SMS settings for the individual. The positive results obtained emphasize the importance of individual fine tuning of all hearing instruments, but particularly those with complex features such as SMS. Thus, the hearing aid fitting should include individual attention and fine tuning of gain, frequency response, compression characteristics, and noise reduction strategies. Proper adjustment of these parameters and fine tuning over time will help ensure maximum benefit to the hearing impaired listener.
|This article was submitted to HR by Matthias Latzel, PhD, Juergen Kiessling, PhD, and Sabine Margolf-Hackl of the Dept. of Audiology at the ENT University Hospital in Giessen, Germany. Correspondence can be addressed to HR, or Matthias Latzel, HNO-Klinik der Justus-Liebig-Universitaet, Funktionsbereich Audiologie, Feulgenstr. 10, 35385 Giessen, Germany; email: Matthias.Latzel(at)hno.med.uni-giessen.de.|