We launched our new hearing aid recording lab this weekend at CES. Check it out:
How about ReSound? You have quite a few other brands that rank lower in sales.
Rather than having to dig through the white paper, it would be interesting to know in summary how the various recordings are scored since that seems subjective. Are there multiple scorers? It also seems if multiple evaluations are done in a row, scoring could be affected by “scorer fatigue,” or even the converse. For example, if one listened to a string of particular bad recordings for several different brands, suddenly hearing recordings from better than average HA’s might cause the scorer(s) to give the better-than-average HA higher marks that it might otherwise obtain if it were reviewed amongst a string of other better-than-average HA’s.
Edit_Update: I see in the YouTube video in the above post that scoring is done according to the following methodology (see transcript quote below), which does not say whether scoring is totally automated or whether any subjective human judgement comes into play - perhaps in how the “leveraging” or “averaging” comes into play?
… We use metrics from the hearing science literature that leverage models of
sensory neural hearing loss to predict the speech intelligibility of each recording. We average
those metrics separately across quiet and loud sound scenes and transform them into our speech
perception benefit metrics. You can explore our content today on hearingtracker.com where
you can audition hearing aids and compare how much they improve speech perception.
Perhaps in critiquing the scoring methodology, one might go to the other extreme and say that actually no human judgement of actual speech quality was involved, only a machine-based set of artificial algorithms
So far very interesting post, I’m looking forward to this being played out, AI or Human, a very subjective subject for most, as for ones hearing as well, but regardless of all of that, I like the “independent” and objective aspect of this, having no manufacturer involved is critical.
I always found that the best way to test aids is to wear them outside in the real world under different situations for at least two weeks. The brain needs time to adjust.
The white paper is short. Check it out. The only subjective part was listening and rating feedback and this was done blindly in a user test. Sounds were presented and the user rated the sound based on whether the audio file presented with mild moderate etc feedback. These ratings were then averaged across users. I was one of the users and the test was legit.
Interesting results! Things I found interesting: Phonak Lumity was rated very high in quiet, but Starkey was rated highest in noise. Phonak was also rated very high in streaming quality. It was also rated very high in feedback management. I have mixed feelings about the feedback score as I have found Sonova aids to reduce a lot of high frequency gain unless they have a fairly occlusive dome/mold. Perhaps the Lumity has improved on this. Although this is interesting and I’m sure there’s high demand for a scoring system for hearing aids, I think people would be better off finding a good fitter.
I was wondering if much of the scoring system is matching “ideal” waveforms. I remember the “robot conversation” topic years ago where the question was, “would you mind a robotic hearing sound if speech recognition is perfect?” I often find that by tuning the relative amounts of bass, midtone, and treble, depending on what noises are in the environment, I can greatly improve speech clarity but can end up with, for example, very tinny sounding voices. I don’t mind the altered sound as it greatly improves my understanding - an example would be a not-so-great podcast in which a guest is Zooming in to a radio show and has a froggy, low-bass rasp in their voice. Turning down the bass and turning up midtones and treble will make the voice a lot clearer (at least to me). I wonder if such an arrangement would score poorly in Hear Advisor rating because it would be off the “ideal” waveform?
I do not grasp the scoring system, but anybody with a high frequency loss is going to initially hear well fit hearing aids a sounding tinny because they’ll be hearing highs they haven’t heard in awhile.
I must have missed it. Where in this site do we see the results of the tests?
Click on the link in the lower part of the OP above. Then when you get to the Hear Advisor page, click on one of the hyperlinks depending on what brand of HA you want to see the results for.
from the www.hearadvisor.com web page:
Hear and Compare our content on devices from: Apple, EarGo, Lexie, Lexie+Bose, Lucid, Jabra, Nuheara+HP, Phonak, Oticon, Signia, Sony, Starkey, and Widex
You should see on each HA review page right under the top “Hear Advisor Lab Ratings” with the rating and the ability to play the recording as the test dummy “heard” it.
Thanks. The transcript said hearingtracker not hearadvisor.
Well, the actual reviews are on HearingTracker.com. One just has to find them in select HA reviews. It would have helped to put the links that I quoted from the hearadvisor.com web page in Abram Bailey’s original post. Maybe Abram might want to update his original post above? - but maybe the Hear Advisor enterprise wants the publicity?!
What’s the easiest way to learn if you’ve tested new models? Thanks.
Good question. We plan to test most of not all new aids we add to HT, primarily the popular ric models at first. But we don’t have a notification for new devices and probably should. Thanks for the suggestion I’ll make it happen.
A dedicated forum thread, usually locked to comments could work too. Pinned whenever it is updated.
Interesting. Love the idea. I took a listen to Apple and Phonak. It appears from the ratings that Phonak Lumity doesn’t do well in noise. Sure would have expected better.
Without my hearing aids, I can hear much of the speech in the Open Ear clip. The 2 clips of the speech with aids sound staticky. These are the busy cafe clips. Given my hearing loss, the first observation surprises me. The second one surprises me because I thought speech was supposed to be clearer “with” aids. I have reverse slope loss–not the more common kind of loss the manikin has–so that surely makes some kind of difference. My hearing loss is also mixed type.
Another thing I noticed, since the verbiage is the same and it’s short–after I heard it a couple of times I knew what I’d be hearing (and maybe filled in rather than really heard it all?)
Did you do any pre-calibrating of the rating scale among the users before doing the ratings for the actual testing? How closely did the multiple raters rate the same sounds? (Inter-rater reliability)
According to their white paper the lead author works for Bose. Says he didn’t directly evaluate any devices associated with Bose.
Steve does everything related to measurements … Also working with Andy (Bose), I can tell you he’s the kind of guy that would ask permission before opening your junk mail. Very principled, by the book, kind of guy…
Feedback test is the only rating that includes subjective ratings.