For anyone who fits hearing aids with directional microphones, the impact of the head-shadow effect and microphone placement on directionality is an important consideration. The practitioner should be aware of the realistic pattern of directionality (including the head-shadow impact) and its interaction with the test set-up when verifying directional benefit via clinical speech testing. And the fitter should be mindful that the specifications provided for directionality are not always reflected in the user's actual experience.
Directionality can be thought of as a decrease in the sensitivity to sound that originates from angles other than the desired angle, usually the front. Directionality can be achieved in hearing instruments by means of a single directional microphone with multiple ports or a system with two or more microphones.
Typically, directional hearing instruments have maximum sensitivity to sound from the front and least sensitivity to sound from somewhere in the rear plane. This pattern of sensitivity, known as a polar pattern, can vary in the rear plane depending on the setting of the microphone(s) and the hearing instrument. Different polar patterns include: cardioid (most attenuation from the rear), bi-directional (most attenuation from the sides), and hypercardioid (most attenuation from the sides and the rear, with a small lobe positioned directly behind).
Polar patterns are displayed as a plot of sensitivity to sounds at different angles for different frequencies, known as polar plots. They are typically generated with pure-tone inputs and the instrument set for flat linear. The plot is a graph displaying the gain or output of a hearing instrument as it is rotated 360 degrees with respect to the noise source. (Alternatively, the noise source can be rotated with respect to the instrument.) The resulting plot displays the amount and pattern of attenuation for sounds coming from different angles.
A polar pattern can be measured in a free field or in situ. For a free-field measurement, the instrument is mounted such that the sound arrives at the microphone completely unobstructed. An in situ measurement is taken with the hearing instrument positioned as worn and is often obtained using a Kemar mannequin in the laboratory.
Certain factors, such as matched microphones in a multi-microphone system, the selected polar pattern, and low-frequency compensation, affect both free-field and in situ polar plots. In addition, with in situ measurements various other factors can also significantly affect the polar plot because the instrument is on the head. These factors include hairstyle, instrument style (i.e., BTE or ITE), microphone placement, and the head-shadow effect. Therefore, the results obtained for the same instrument in a free field compared to in situ can differ significantly.
A potential advantage of digital dual-microphone systems (with two analog-to-digital converters) is that multiple patterns of directionality can be applied to the same instrument.1 If the delay between the microphone responses is changed, the instrument can produce different polar patterns. The clinician can select these varying polar patterns through the fitting software, which in turn programs the delay between the microphones.2
Another complicating factor is that ambient noise frequently changes in intensity and location. To provide an optimal directional pattern at all times, an adaptive directionality (AD) algorithm can be used. This algorithm continually monitors the environment and can change the direction of the null in the rear plane of the polar pattern up to 250 times a second. This technique determines what sound is coming from the front and what from the sides, the back, and in between.
To determine what is coming from the front, the instrument uses the AD algorithm to analyze the output of each microphone with respect to the front microphone. It does the same with respect to what is coming from the back microphone. Then, it uses an adaptive filter to maximize the noise cancellation by automatically changing the filter coefficients according to a noise-minimization algorithm. AD places the null of the microphone sensitivity at the site of the most intense noise source originating from the side or rear hemispheres. Potential advantages of AD are to ensure the best polar pattern for any given situation and to address the head-shadow effect.
It has been demonstrated that effective directional systems provide the listener with an improved signal-to-noise ratio (S/N) for sounds coming from the front.3,4 However, the amount of benefit achieved by the listener, as measured by clinical speech testing, depends on the instrument style (microphone placement), head-shadow effect,5,6 and the test set-up (speaker placement).7,8 Furthermore, adaptive directionality also affects the results because the polar pattern changes to accommodate the test set-up.
This article will demonstrate the effect of head shadow and microphone placement on polar plots, as well as the interaction between these effects and subject performance in noise. We conducted three separate investigations for this study: (1) a comparison of free-field and in situ polar patterns; (2) a comparison of benefit obtained by subjects wearing a hypercardioid and an adaptive directional polar pattern; and (3) a comparison of real-ear aided gain (REAG) measurements of an ITE and a BTE with varying polar patterns.
INVESTIGATION 1: POLAR PLOTS
We obtained polar plot measurements for omnidirectional, hypercardioid, and cardioid patterns at 1000 and 2000 Hz for GN ReSound digital BTE and ITE instruments, both in a free field and in situ on KEMAR. The instruments were set for a flat 20 dB of linear gain.
Results and discussion
Figure 1 displays free-field polar plots measured for the digital BTE instrument at 1000 Hz for omnidirectional, hypercardioid, and cardioid polar patterns.
The style of instrument (BTE versus ITE) has minimal effect in the free field and there is little difference between 1000 and 2000 Hz in the free field. However, when an instrument is measured in situ, the microphone placement and head-shadow effect produce a dramatic effect, as is shown in Figures 2–4.
Figure 2 displays what happens to the polar pattern with an omnidirectional microphone setting. Even before a directional algorithm is applied, the responses of the instruments (particularly the ITE) are already attenuated for signals originating from the opposite side and from behind, due solely to the microphone placement and head-shadow effects. Figures 3 and 4 show the additional effect on an already altered response when a directional pattern is added to the processing. A close look at the ITE in situ polar patterns measured for a hypercardioid setting indicates that the head-shadow effect on Kemar now results in a more cardioid-like response for the ITE. Meanwhile, the BTE in situ polar patterns measured for a cardioid setting indicate that the head-shadow effect on Kemar results in a hypercardioid-like response (null shifted to the side) for this BTE instrument. Although the BTE results are less dramatic than the ITE results, head shadow and microphone placement do affect them.
INVESTIGATION 2: REAL-EAR MEASURES
We obtained real-ear aided gain (REAG) measures on Kemar, again using GN ReSound digital BTE and ITE instruments. The noise stimuli were presented at 0º and 180º azimuth with the instruments programmed with omnidirectional, cardioid, hypercardioid, and adaptive directional patterns. The instruments were set for 1.5:1 compression with 20 dB of gain for a 50 dB SPL input, and 10 dB of gain for an 80 dB SPL input. The stimulus was a 65 dB SPL composite (speech-weighted noise) signal delivered by the Frye Fonix 6500CX.
Results and discussion
The three polar patterns for the ITE and BTE produced essentially the same result for the 0º condition; therefore, we have displayed only one response for the gain at 0º. Results indicate that the ITE and BTE instruments set for the same polar pattern respond with different amounts of attenuation at 180º. This supports the results of the effect of head-shadow and device style seen on the in situ polar plot measurements.
In Figure 5, the REAG for the BTE instrument shows slightly more attenuation for the cardioid polar pattern than for the hypercardioid pattern at 180º relative to 0º. As with the in situ polar patterns, the gain attenuation from the rear between fixed patterns demonstrates slightly more attenuation with the cardioid pattern as well. Furthermore, the REAG for the BTE instrument shows greater attenuation for the adaptive pattern than for either of the fixed patterns.
Figure 6, which illustrates the REAG for the ITE, demonstrates that the REAG for the hypercardioid has greater attenuation for the signal at 180º than for the cardioid pattern. The in situ polar plots and real-ear measures are consistent with each other; however, the real-ear measures do not reflect the free-field polar plots. Interestingly, the ITE has the same amount of attenuation for the hypercardioid and the adaptive directional pattern.
INVESTIGATION 3: SPEECH TESTING
Part 1: Hearing in Noise Test
Ten subjects with sensorineural hearing-impairments ranging from mild to severe participated in the speech-in-noise testing. Of these, nine were experienced users of binaural digital amplification and one of monaural digital amplification. Three subjects were wearing GN ReSound digital ITE instruments and seven wore GN ReSound digital BTE instruments. All instruments had omnidirectional and multiple-pattern directionality capabilities.
We used the Hearing in Noise Test (HINT) to determine the benefit of hypercardioid and adaptive directional patterns over an omnidirectional pattern. The HINT is an adaptive procedure in which the masker (uninterrupted speech-shaped noise) remains at 65 dBA and the speech signal (sentences) varies.9 The sentences were delivered at 0ºazimuth and the noise was delivered at (1) 110º and 250º, (2) 180º, and (3) 110º, 180º, and 250º.
Subjects were tested in three conditions: (1) omnidirectional, (2) fixed hypercardioid directionality, and (3) adaptive directionality. The HINT score is reflected as a signal-to-noise ratio (S/N) threshold. A lower S/N threshold suggests better performance, indicating that the subject could understand speech in a more difficult listening environment. Furthermore, the S/N thresholds were compared between conditions to yield a benefit score.
Results and discussion
Figure 7 displays the mean S/N scores for omnidirectional, fixed hypercardioid (FD), and adaptive directionality for various noise conditions. A Wilcoxon Signed Ranks Test (non-parametric) indicated that for each noise condition there was a significant difference in performance (p<.028) with both fixed hypercardioid (FD) and AD directionality over the omnidirectional program.
Of particular interest for this speech test are the S/N differences between adaptive and fixed directionality and between device styles (ITE vs. BTE). Results demonstrated that 8 of the 10 subjects performed significantly better (>1.5 dB difference between conditions) with AD than with FD when speech was at 0º and noise was at 180º. This result is expected because in this noise condition, the adaptive pattern will direct the null toward the noise source at 180º, while the fixed hypercardioid pattern maintains its lobe at 180º. In the more diffuse noise conditions, we observed very little difference between the two types of directionality. This was expected because the adaptive directionality will default to the “optimal” hypercardioid pattern in a diffuse noise environment.
A closer look at the 180º noise condition revealed that the two subjects who did not achieve significantly greater benefit with AD over FD were both wearing ITEs. Figure 8 shows the HINT results of adaptive and fixed hypercardioid (noise at 180º) separated by instrument style. These results suggest that device style (microphone placement) and head-shadow effects can affect the benefit achieved with different polar patterns (hypercardioid vs. adaptive). Needing more data on ITE users to verify this finding, we then conducted additional speech testing.
Part 2: Connected Speech Test
We conducted this investigation on a separate group of subjects wearing the same type of binaural ITE instruments with omnidirectional and multiple pattern directionality capabilities. We used the Connected Speech Test (CST) to evaluate 14 subjects (28 ears) with mild to profound sensorineural hearing loss.
The CST is a sentence recognition test in which subjects are visually given a topic word and then instructed to repeat sentences on that topic presented in the sound field. For each condition, 4 lists of 10 sentences with 25 key words are presented. Scoring is based on percentage of key words correctly identified.10
Sentences were always presented at 60 dBA from 0º azimuth in a sound-treated booth. The six-talker babble was always presented at an overall level of 60 dBA (0 dB signal-to-babble ratio). Each subject completed 4 lists per condition and 10 sentences per list (6 conditions, 24 lists, 240 sentences) in randomized order. Subjects were tested wearing: (1) an omnidirectional pattern, (2) a fixed directional hypercardioid pattern (FD), and (3) an adaptive directional program. Furthermore, each of these conditions was tested with noise sources at 110º, 180º, and 250º (FD3 and AD3) and with the noise source at 180º only (FD1 & AD1).
Results and discussion
A Wilcoxon Signed Ranks Test indicated that subjects performed significantly better (p<.001) with both directional programs (FD & AD) than with the omnidirectional program (19%-23% improvement in recognition). There was no significant difference between the two types of directionality (FD vs. AD) in either noise condition (one source or three sources in the rear plane). These CST results are consistent with the HINT results for the ITE wearers.
When the noise source is located at 180º, the AD algorithm shifts the null to 180º. As seen in the polar plots for ITE style instruments, the head-shadow effect and microphone form a “natural” null around 180ºfor high frequencies. Therefore, for ITE instruments, the AD's shifting the null to 180º does not improve the already existing reduction in gain at 180º created by the head-shadow effect coupled with the hypercardioid response.
Speech data correlate well with and may be explained by in situ polar plots and REAG measured on Kemar. These objective data support the difference seen in HINT benefit scores between BTE and ITE style instruments for the 180º noise condition. Additionally, the real-ear data reflect the differences seen in subject performance between an adaptive and hypercardioid directional pattern at 180º. All three investigations help explain the interaction between device style, head-shadow effect, adaptive and fixed directionality, and subject performance on speech-in-noise tests.
When comparing AD to a fixed polar pattern, one must consider the “optimal” polar pattern for the noise condition. When noise is coming from directly behind (180º), the adaptive pattern will default to a cardioid-like pattern for optimal noise suppression. Although during speech testing in these investigations, both the BTE and ITE instruments were set for hypercardioid patterns, the polar plots and REAG indicate that a BTE responds like a hypercardioid when worn in situ, but the ITE instruments respond more like a cardioid when worn in situ. Therefore, with noise at 180º, the adaptive pattern (cardioid-like) provided benefit over hypercardioid for the BTE instruments (responding like hypercardioid in situ), but not for the ITE instruments (responding like cardioid in situ).
Depending on the individual's head shape, a single noise source located at a slightly different azimuth in the rear plane may demonstrate the benefit of adaptive over fixed directionality for an ITE instrument. For example, with the noise source at 90º, AD would provide the greatest attenuation of that noise.
A further test could be designed to demonstrate the reverse of this scenario, one where adaptive and fixed directionality perform the same for BTE instruments, but AD performs better for ITE instruments. The outcome would depend on the polar pattern selected, the instrument style, and the location of the noise sources.
In our studies, real-ear, HINT, and CST results all demonstrated that AD performed as well as or better than any of the fixed directional patterns.
It is important to consider the head-shadow effect and instrument style when analyzing directional benefit for any given polar pattern in any given noise condition. Although free-field measurements depict exactly how the directionality in an instrument is performing, it is the objective, in situ measurements that demonstrate the polar pattern that the patient will actually receive. As demonstrated in this study, the subjects' speech data correlate well with the polar plot and REAG measurements on Kemar.
The findings reported here show that a fixed polar pattern can act very differently from expected when it is worn on a head, due to head-shadow effect and microphone placement. This finding suggests that a polar pattern will respond differently on differently shaped heads and with different instrument styles, and further investigation is warranted to examine these issues.
In all of the data presented here, the adaptive directionality pattern performed as well as or better than any of the fixed directional patterns. It follows, therefore, that AD may be useful in providing an optimal directional pattern for varying head shapes, microphone locations, and noise conditions.
The authors would like to acknowledge the contributions of Srdjan Petrovic, senior electroacoustic engineer, GN ReSound Group.