Today, the most successful way to improve speech intelligibility in noise with hearing instruments is by the use of directional microphones. With the use of directional hearing instruments, noise from behind or the sides is reduced, the signal-to-noise ratio is improved and speech understanding in situations where speech and noise are spatially separated is significantly enhanced. A multitude of studies has shown the directional benefit to the hearing impaired.1-3

There are two main approaches to implementing directivity in hearing instruments. One is the “omni-plus-directional” approach, which uses one omnidirectional microphone and one directional microphone with the possibility of switching between them, so that only one microphone is activated at a time. The second approach is dual microphone implementation, in which both microphones are omnidirectional. An electronic delay is introduced between the output of the two microphones to achieve directivity. If omnidirectional characteristics are desired, only one of the two microphones is activated.

One advantage of the dual microphone approach compared to omni-plus-directional implementation is its flexibility. While conventional directional microphones are restricted to one specific polar pattern, the dual microphone approach allows for electronic adaptation of the polar pattern, depending on the acoustical situation. When noise comes from the back, a cardoid pattern can be applied, and when the noise comes from the side, a hypercardoid pattern provides adequate attenuation. The directivity pattern can be automatically altered according to the specific acoustic situation. Evidence suggests that hearing systems with automatic adaptive directionality outperform systems with fixed directionality when the noise comes from the side,4 or in situations where the noise source is moving or diffuse (i.e., situations of high relevance in daily life).5

Microphone Drift and Debris
For optimal performance of the dual microphone approach, however, the respective sensitivities of the two microphones across frequencies have to be well matched. A mismatch between the two microphones can severely reduce the directivity of the system. There are different sources of mismatch, each affecting directivity differently:

Microphone mismatch due to aging (referred to as “drift”) is caused by a parallel shift of the frequency responses between the two microphones. This type of mismatch can be compensated for with a frequency-independent adjustment of the gain for one microphone.

Microphone mismatch due to dirt or moisture in the acoustical pathway (e.g. in the microphone port), which causes a frequency-dependent modification of the frequency response (often a reduction in the high frequency region). This second type of mismatch can only be compensated for by a frequency-dependent adjustment of the gain.

The following study examines the impact of both types of microphone mismatch on directivity, and the consequences on speech intelligibility.

Microphone Mismatch: Effects on Directivity
Aging: Fig. 1 shows the average front-to-back ratio from 100 Hz to 10 kHz, as a function of microphone drift, that can occur due to aging.

chart
Fig. 1. Average front-to-back ratio (attenuation in dB) between 100 Hz and 10 kHz as a function of microphone drift. With well-matched microphones (0 dB drift), the back direction is attenuated by -11 dB, compared to the front direction.

The function was measured with a Phonak Claro 211 dAZ BTE instrument, which uses dual-microphone technology. The microphone drift was artificially generated by software. With well-matched microphones (0 dB drift), the signal from the back direction (a pure-tone sweep) is attenuated by -11 dB on average, compared to the signal from the front direction. A drift (meaning a difference in broadband sensitivity) of ±0.25 dB only slightly affects directivity, but with further increasing drift, directivity decreases.

Debris: The second source of microphone mismatch is dirt and debris in the acoustic pathway. The influence of dirt on directivity was measured with 13 of the instruments described above, which were in use by customers for about one year. A replaceable, acoustically transparent filter protects the microphones of these instruments against dirt and humidity. Prior to the measurements, age-related drift was measured and compensated for (average microphone drift after one year: 0.25 dB), so that the influence of dirt could be examined independently. Fig. 2 shows the average front-to-back ratio of the instruments with used filters, and with replaced, clean filters as a function of frequency.

chart
Fig. 2. Average front-to-back ratio (attenuation in dB) across frequency with used microphone covers, and with replaced covers.

It can be seen that directivity is increased after changing the filters, especially in the mid- and high-frequency region. However, there were large variations between the instruments used by the individuals. Some of them showed no differences after filter replacement; the directivity of other instruments strongly improved after filter replacement, especially those with clearly visible signs of soiling.

Effect on Speech Intelligibility
The measurements described above illustrate the reduction of directivity with microphone mismatch due to microphone aging and dirt in the acoustical pathway. To examine the impact of mismatch on speech intelligibility, tests were conducted with hearing-impaired and normal-hearing listeners.

The six hearing-impaired subjects had moderate to severe sensorineural hearing losses. Speech tests in noise were conducted with the Oldenburg Sentence Test.6 Speech was presented from the front (0°), noise (party babble) from the back (180°). The speech tests were conducted in an audiological examination room. The SNR of the speech reception threshold (SRT: 50% speech intelligibility) was determined using an adaptive procedure. The subjects were monaurally fit with Phonak Claro 211 dAZ instruments, and the unaided ear was blocked during the measurements.

Aging: The SRT was measured in the directional mode with matched microphones, and with artificially generated microphone drift ranging from –1.5 to 1.5 dB. In addition, the SRT was determined in omnidirectional mode. The results for the hearing impaired subjects are plotted in Fig. 3.

chart
Fig. 3. Mean SRT results for hearing impaired subjects in directional mode as a function of microphone drift, and in omnidirectional mode (right). The lower the SRT value, the better the performance.

It can be seen that microphone drift only had a limited influence on the performance of the subjects. The mean SRT was -8.7 dB for well-adjusted microphones, and deteriorated by approximately 1 dB towards strong microphone mismatch. This corresponds to a speech intelligibility reduction of about 17 percentage points.7 Even with de-adjusted microphones, speech intelligibility in the directional mode was clearly improved, compared to omnidirectional mode.

For comparison, the same experiment was conducted with five normal hearing subjects. They were fit monaurally with the above described DSP instruments using unvented earmolds or soft plugs. The unaided ear was blocked during the measurements. The mean results are shown in Fig. 4.

chart
Fig. 4. Mean SRT results for 5 normal hearing subjects in directional mode as a function of microphone drift, and in omnidirectional mode (right).

Similar to the results for the hearing-impaired group, a microphone drift in the range of ±0.5 dB does not affect speech intelligibility. With increasing drift, a degradation of speech intelligibility can be observed.

The large standard deviation between ±0.5 dB microphone drift can be explained by the varying level of speech test experience between the subjects. Two subjects had no previous experience with the Oldenburg sentence test. Like the hearing-impaired subjects (who also had no experience with the test), their performance was only slightly correlated with microphone drift (if at all). The three remaining subjects, in contrast, were very familiar with the test and its speech material. Due to their extensive training, they were able to perform “at the limit” and reached very low SRT values. Hence, they could detect even small changes of the acoustic situation. Between –0.5 and 0.5 dB microphone drift, they showed no degradation of speech intelligibility. With more severe drift, however, understanding decreased in this sub-group.

Debris: To examine the effect of dirt in the acoustic pathway on speech intelligibility, the five normal hearing subjects were each fit with three of the above described DSP instruments with used filters. Again, the Oldenburg sentence test was measured under the same conditions. The average SRTs in directional and omnidirectional mode are shown in Table 1.

    HI #1 HI #2 HI #3 Omnidirectional
SRT [dB] -17.3 -14.2 -11.6 -10.9
Table 1. Mean SRT with 3 used instruments in directional and omnidirectional mode (five normal-hearing subjects). Dirt in the acoustical pathway can almost entirely neutralize directivity (hearing instrument #3).

With instrument #1, the average SRT was at –17.3 dB SNR, which is comparable to the results measured with a new instrument. (See Fig. 4. The average SRT with matched microphones in directional mode across all five normal hearing subjects was –16.6 dB.) With instrument #2, in contrast, speech intelligibility is clearly reduced. With instrument #3, the effect of directivity is almost completely neutralized, and speech intelligibility is roughly the same as in omnidirectional mode. By visual inspection, it was obvious that the microphone cover of this instrument was soiled.

These results show that dirt in the acoustic pathway can have a severe impact on directivity and speech intelligibility in noise. Automatic or manual microphone matching which is frequency-independent cannot compensate for this effect. In the tested hearing instruments, however, a simple replacement of the microphone covers is sufficient to completely restore directivity.

Summary
There are two sources of microphone mismatch in dual microphone systems, which should not be confused: age-related microphone drift, and mismatch due to dirt in the acoustical pathway. The effect of age-related drift seems to be limited. The data collected so far indicate that the average drift is about 0.25 dB after one year, which is unlikely to affect the directional benefit for the hearing impaired. As already pointed out by Csermak: “Regardless, no evidence to date has been produced which suggests that any drift that might occur is sig-nificant enough to have a noticeable effect.”8

Automatic compensation for age-related drift which is frequency independent and matching the microphones during use is possible by monitoring and minimizing the long-term overall level difference between the microphones. However, there is the potential danger of further mismatching the microphones if the shape of the two microphone response curves differ due to dirt in the acoustic pathway. Compensating for overall level differences could introduce “de-adjustment” in frequencies which are important for speech intelligibility.

The experiments indicate that dirt in the acoustic pathway can have a strong impact on directivity and speech intelligibility in noise. In some cases, dirt almost entirely neutralized directivity. Thus, it is important to carefully protect the microphones from soiling. An automatic, frequency-independent adjustment of the microphones cannot compensate for microphone mismatch due to dirt, as this is a frequency-dependent effect.

Filters have the capability of providing effective protection of the microphones against dirt and moisture. They are easy and inexpensive to replace, if necessary by the customer him- or herself. If the microphones are not sufficiently protected against dirt and moisture, expensive microphone replacement is necessary to restore directivity.

This article was submitted by Jürgen Tchorz, PhD, an audiologist in the audiological communication and education department at Phonak AG, Stäfa, Switzerland. Correspondence can be directed to HR or Jürgen Tchorz, PhD, Phonak AG, Laubisrütistr. 28, 8712 Stäfa, Switzerland; email: [email protected].

References
1. Ricketts T & Dhar S: Comparison of performance across three directional hearing aids. J Amer Acad Audiol 1999; 10: 180-189.
2. Thompson SC: Dual microphones or directional-plus-omni: Which is best? In S Kochkin & KE Strom’s (eds) High Performance Hearing Solutions, Vol 3, Suppl to the Jan 1999 Hearing Review: 31-35.
3. Valente M: The bright promise of microphone technology. Hear Jour 1998; 51 (7): 10-19.
4. Checkley P & Kühnel V: Advantages of an adaptive multi-microphone system. Hearing Review 2000; 7 (5): 58-60, 74.
5. Ricketts T: Brit Jour of Audiol (submitted).
6. Wagener K & Kühnel V & Kollmeier B: Entwicklung und Evaluation eines Satztests in deutscher Sprache I: Design des Oldenburger Satztests. Zeitschrift für Audiologie/ Audiological Acoustics 1999; 38 (1): 4-15.
7. Wagener K, Kühnel V & Kollmeier B: Entwicklung und Evaluation eines Satztests in deutscher Sprache III: Evaluation des Oldenburger Satztests. Zeitschrift für Audiologie/Audiological Acoustics 1999; 38 (3): 86-95.
8. Csermak B: A primer on a dual microphone directional system. Hearing Review 2000; 7 (1): 56-60.