International Speech Test Signal and Feedback
Background
The International Speech Test Signal (ISTS) was developed to determine how speech is processed by modern hearing aids {{1}}[[1]] Holube, I., Fredelake, S., Vlaming, M. & Kollmeier, B. (2010). Development and analysis of an international speech test signal (ISTS). International Journal of Audiology, 49, 891-903[[1]]. The signal is based on natural recordings, but is largely non-intelligible because of segmentation and remixing. When used for hearing aid and probe-microphone measurements, the signal describes the gain at different percentiles of the speech level distribution (roughly the 30-dB dynamic range of speech). The ISTS was developed by the ISMADHA working group within EHIMA (European Hearing Instrument Manufacturers Association) and made available to all interested in the measurement of hearing aids.
It occurred to me that the ISTS might lend itself to assisting with a basic hearing aid fitting problem: determining the impact of acoustic feedback in real-life-like settings rather than relying only on the programmed settings provided by the manufacturer, and then waiting for the patient’s response after leaving the office and gaining some real-life experiences. While the programmed settings attempt to manage feedback by adjusting various parameters, especially at levels where feedback is expected to occur, patients still often experience feedback under certain listening circumstances. Could the determination of feedback, as a problem, be determined before the patient leaves the office? Could the ISTS provide a predictor of acoustic feedback for normal listening conditions?
Method
To check this out, three closed coupled BTE RIC (receiver-in-the-canal) hearing aids were evaluated. Hearing aid “A” was a lower-gain adaptive AGC (automatic gain control with a wide dynamic range compression) hearing aid, and hearing aids “B” and “C,” higher-gain, essentially linear, instruments. Hearing aid “A” was programmed for three different responses, while hearing aids “B” and “C” were programmed for two similar linear responses, differing primarily in gain. The instruments were measured according to ANSI ’03, and then with ISTS. Measurements were made using a Frye Electronics Inc. model 8000 hearing aid test system for both measurements.
Visual comparisons of the ANSI and ISTS measurement methods were made to determine if anything in the response curves would suggest susceptibility to acoustic feedback. ISTS measurements were made using a 50-dB SPL input to provide consistent gain comparisons with the 50-dB SPL input signals used for ANSI measurements. This was especially true for the WDRC (wide dynamic range compression) hearing aid (Aid “A”), because, as designed, it should have greatest amplification for low-level input signals. Therefore, it seemed logical that it would be at this level that feedback should be most obvious. This would not be true for linear hearing aids.
In some cases, additional input levels were used for ISTS measurements to determine the impact of changes in the input signal to the recorded results. Figure 1 provides the color key for ISTS curves that are shown under the Results Section in this blog. It is important to keep in mind that the percentiles for the ISTS are referenced to the LTASS (long-term average speech spectrum). For interpretation of the ISTS graphs, the following explanation is given for the graph colors:
- Green = LTASS (Long-term average speech spectrum)
- Blue = 99th percentile
- Red = 65th percentile
- Yellow = 30th percentile
Figure 1. Colors identifying the various percentiles of the ISTS CURVES. Green – LTASS; Blue – 99th percentile; Red – 65th percentile; Yellow – 30th percentile.
Results
Figure 2 compares the ANSI and ISTS for setting #1 of hearing aid “A.” A question might be asked about the irregularities in the ANSI gain response from about 3 to 5 kHz (2a). Does this suggest susceptibility to feedback? Based on the measurement alone, it is impossible to tell. The ISTS measurement (2b) suggests that feedback would not be an issue because none of the 1/3 octave bands show a significant deviation from the overall response between the 30th and 99th percentiles. (In comparing the ANSI and ISTS gains, recall that the ANSI SPL curve representing the Response [green curve] has to have 50 dB subtracted from it to provide the gain value).
Figure 2. The ISTS (2b) follows the ANSI (2a) response curve fairly well. None of the 1/3-octave bands shows any tendency toward feedback, regardless of the percentile curve.
Figure 3 compares the ANSI and ISTS for setting #2 of hearing aid “A.” This is a high-frequency response setting that might be suspected of being prone to feedback. However, at its maximum gain setting (50 dB SPL input), the ISTS gives no indication of feedback at any of the percentiles. Greater gain is shown at the 30th percentile, but that would be expected due to the hearing aid’s greater gain response to the softer portions of the test signal. The LTASS is fairly close to the ANSI gain curve.
Figure 3. The ISTS (3b) follows the ANSI (3a) response curve fairly well. None of the 1/3-octave bands shows any tendency toward feedback, regardless of the percentile curve.
Figure 4 shows the results of hearing aid “A,” setting #3. The ANSI gain response curve (4a) has a couple of curious peaks in the higher frequencies, but does not look too unusual. However, when the same aid is tested using the ISTS (4b), a definite peaking of energy is detected in one of the higher 1/3 octave bands – that around the 4000 Hz area. So, at least for soft input signals, the hearing aid would be characterized with feedback. However, when an 80-dB input signal is used (4b inset), the ISTS shows no feedback. This would be expected, especially since the gain is reduced by the hearing aid as the input signal increases. Still, this would appear to be a problem hearing aid fitting, and in this case it would appear that the ISTS provides significant information, at least as related to feedback.
Figure 4. The ISTS (4b) shows an exaggerated peak around 4000 Hz that is not evident in the ANSI gain response (4a). The inset in 4b shows the change in amplification at an 80 dB SPL input, reflecting the reduced gain with higher input levels, and hence, no feedback.
Figure 5 shows the gain response and ISTS response for hearing aid “B,” a high-gain RIC hearing aid. The gain curve in Figure 5a suggests acoustic feedback, and this is conformed in Figure 5b with all percentiles showing excessive amplification around 3500 Hz, along with some unusual activity at higher frequencies.
Figure 5. The ISTS (5b) shows an exaggerated peak around 3500 Hz that would seem to correlate with a similar peak in the response curve (5a).
Figure 6 shows the results of hearing aid “B” programmed somewhat differently. In this case, the ISTS (6b) is very similar to the gain response curve (6a), and no exaggerated 1/3 octave peaks exist to suggest feedback. The overlaying of responses for the different percentiles would be characteristic of a linear amplifier, where gain is constant.
Figure 6. The gain curve (Fig. 6a) is very similar to the ISTS curve (Fig. 6b). No feedback is suggested by either of the graphs.
Figure 7 is hearing aid “C” programmed for high gain. The ANSI peak gain (7a) suggests a feedback problem. The ISTS response (7b) confirms this.
Figure 7. The gain curve (Fig. 7a) suggests the possibility of feedback. The ISTS confirms this (Fig. 7b).
Figure 8 (hearing aid “C”) has been programmed for fairly linear gain characteristics. The ANSI gain response (8a) might cause one to suspect that feedback might be a problem. However, the ISTS response (8b) suggests that the gain peak at around 3000 Hz is actually that – gain.
Figure 8. The gain curve (Fig. 8a) appears to have a potential peak problem around 3000 Hz. However, the ISTS (Fig. 8b) does not show this to be a feedback problem, but the basic response of the hearing aid.
Summary
Comparisons of ANSI and ISTS measurements on a variety of RIC hearing aids suggests strongly that the ISTS could be used as a simple method to better fit hearing aid users with respect to feedback management, along with other considered advantages of the ISTS.
