When a new algorithm requires a specific test: the story behind our live feedback test

Reading Time: 6 Minutes
08.05.2019

Testing is a crucial step while developing our products as it allows us to select the most suitable solution. How can we evaluate a defined algorithm in specific situations? This blog post will introduce you to the process of developing a test to evaluate our new Dynamic Feedback Canceller (DFC) in dynamic situations. The resulting live feedback test has been designed to evaluate feedback cancellation systems in situations reflecting everyday life. We’ll also share the results with you.

Acoustical feedback and hearing aids: the same old story

While market surveys like MarkeTrak or EuroTrak report increasing satisfaction in hearing aid use over the years, handling acoustical feedback remains a top priority for hearing aid users. There is no need to engage extensive scientific resources to understand that unwanted whistling is not acceptable for the hearing aid user and his entourage. Acoustic feedback might impact intelligibility beyond the embarrassment of having an identifiable and malfunctioning hearing aid. Providing a feedback-free fitting ensures benefit for the end user but also for the audiologist with less constraints regarding amplification and by keeping the acoustical coupling as open as possible. This is a motivation for our R&D team to constantly improve the feedback canceller in our hearing aids.

Static and dynamic conditions

Acoustical feedback might occur in two different conditions when fitting a hearing aid. It might occur spontaneously if gain is programmed above the feedback threshold or if the acoustical coupling allows too much leakage from the speaker to the microphone. It can also arise when you check the sensitivity to feedback in dynamic conditions by covering the hearing aid with your hand.

These dynamic situations are illustrated in the videos below where a device is suddenly covered with my hand without and with DFC. The devices are fitted with an open dome and gain set 6 dB above the feedback threshold. It is not a recommended fitting as the systems are especially instable and stressed in these conditions.

Trigger feedback without DFC

Trigger feedback with DFC


The videos clearly differentiate the static from the dynamic conditions. Without DFC, there is a strong acoustical feedback only when I insert the device and each time I cover the ear. However, the hearing aid is stable without any movements around the ear. The device is much more stable during all the manipulations when DFC is activated.

Dynamic Feedback Canceller

Static and dynamic situations are addressed by Bernafon’s DFC™ encompassing two processing blocks. A first block will have a slow estimator of the feedback path. The estimator must be slow to be accurate and precise so that the system cancels only what is needed. This should ensure a good sound quality. However, when the feedback path suddenly changes, then you have a mismatch between the estimated and real feedback path. This might trigger feedback until the new real feedback path is correctly estimated. These fast changes can be detected by specific estimators located in a second block designed to break the feedback loop as early as possible.

A first prototype was ready to be tested once the system was developed. We found out that available test methods like maximum stable gain or aided gain before feedback show limitations in correctly evaluating DFC. They are usually used to compare different feedback cancellation approaches by slowly increasing the gain with a 1 dB step (Spriet et al. 2010, Macrum et al., 2018). A slow gain update without any changes in the feedback path will only test the ability to estimate and cancel feedback in a static situation. These static situations represent only one aspect of feedback cancellation that are handled by the slow and precise estimator of DFC.

Evaluating feedback in dynamic situations

The dynamic situations cover another aspect of feedback cancellation as they reflect realistic situations of the user’s daily life, for example, hugging a relative, taking a phone call, or putting a hat on. Unfortunately, there is no available test that evaluates feedback cancellation with a focus on their dynamic responses and their perceptual consequences. Based on the input from the development engineers and our clinical experience, we quickly identified the need to design a specific test to evaluate the dynamic response of DFC. These dynamic situations should not be made by a robot but should encompass real manipulations to introduce sudden changes in the feedback path.

The resulting live feedback test was designed to test any feedback cancellation system with situations known to trigger feedback, i.e. covering the ear with the hand, the phone, wearing a hat, or while inserting and removing the device. One interesting aspect of the test is to repeat the manipulations with changes to the speed or the amplitude of the movement. For each manipulation, the feedback annoyance is rated on a visual analog scale. A series of manipulations is performed with a device without any feedback cancellation to calibrate the subjective rating scale of annoyance.

The evaluation can be repeated with each system fitted at different gain levels. The idea is to fit the devices at or above the measured feedback threshold to increase the likelihood of seeing a difference between different solutions. Our results (Guo et al., 2018) show that the annoyance of perceived feedback is increased when more gain is added above the feedback threshold. This result is important as it indicates that the test results correctly reflect a change in test condition.

Test results associated to experiences from take-home trial

We used this live feedback test in a clinical trial once the psychometric properties of the test were validated. The test was single blinded and performed in a lab setting with 22 experienced hearing aid users with a moderate to severe hearing loss. The lab test results were completed with subjective outcomes after a 2-week take-home trial with Zerena, as the reference device, and Viron, as the test device with DFC activated.

During the live feedback test, feedback was detected in 52 % of the manipulations with the reference device and in 35 % of the manipulations with the test device. Once feedback occurred, it was rated as being significantly more annoying without DFC. The take-home trial should directly reflect the feedback experience in situations typically experienced by hearing aid users. Results were given on an ordinal scale (never, rarely, sometimes, often, and always) regarding the frequency of feedback events. The results from this 2-week take-home trial indicate a significant reduction in feedback frequency with DFC. This outcome is important because it confirms the external validity of our live feedback test.

Benefit from the lab to your customers

The external validity of a test can be seen as the potential generalization of results outside the test context. As audiologists, we should always try to find a link between results from research and how they relate to our hearing-impaired customers. However, it is often challenging because reality is highly complex, and research only captures a defined and limited aspect of it.

This is especially the case if you only test a feedback canceller in static situations. These findings might be interesting, but they don’t reflect the true usage of a hearing aid. We found evidence that our live feedback test is able to complete the evaluation of different feedback canceller approaches. The results are tangible and directly linked to reported subjective experiences

All the methodology can be found in our White Paper if you are interested in reading more, reproducing our results, or in extending them to your own test cases.

References

Guo, M., Kuriger, M., Lesimple, C., & Kuenzle, B. (2018). Extension and Evaluation of a Spectro-Temporal Modulation Method to Improve Acoustic Feedback Performance in Hearing Aids. In 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE.

Marcrum, S. C., Picou, E. M., Bohr, C., & Steffens, T. (2018). Feedback reduction system influence on additional gain before feedback and maximum stable gain in open-fitted hearing aids. International Journal of Audiology, 57(10), 737–745.

Spriet, A., Moonen, M., & Wouters, J. (2010). Evaluation of feedback reduction techniques in hearing aids based on physical performance measures. The Journal of the Acoustical Society of America, 128(3), 1245

 

About the author:

Christophe Lesimple
Christophe Lesimple
Christophe is a Clinical Research Audiologist and has worked for Bernafon since 2011. He contributes to various aspects of development like working on concepts, running clinical trials, and analyzing data. Besides his activities with Bernafon, he teaches research methods and statistics at the University of Lyon. In his private time, Christophe likes to play music and volunteer for a hearing impaired association.

 

Comments