Audio Amplifier Speech Understandabilility

Would this be of any help?

http://www.eurekalert.org/pub_releases/2009-08/afot-htw080509.php

...Prof. Miriam Furst-Yust of Tel Aviv University's School of Electrical Engineering has developed a new software application named "Clearcall" for cochlear implants and hearing aids which improves speech recognition for the hard-of-hearing by up to 50%.

"Hearing-impaired people have a real problem understanding speech," says Prof. Furst-Yust. "Their devices may be useful in a quiet room, but once the background noise levels ramp up, the devices become less useful. Our algorithm helps filter out irrelevant noise so they can better understand the voices of their friends and family."

Based on a cochlear model that she devised, the new patented technology is now being developed to improve the capabilities of existing cochlear implants and digital hearing aids. Adding Clearcall to current technology is quite straightforward, says Prof. Furst-Yust, and requires only add-on software for existing devices...

I would have thought the hearing aid manufacturers world wide must know this surely.

They must be more than a glorified graphic equaliser in these times of DSP.

As you suggested its probably more a matter of seperating these consonant sounds from the background noise.

My nephew picked up a new (top of the line) hearing aid a month or 2 ago, had it back for adjustment several times. Finally realised it was the sound of the computer fan in the room was drowning most other things out.

I intend to make an instrument or two to help him out passing the last area of his mechanical apprenticeship he needs to get his mechanics papers .. namely 'diagnosing engine faults by ear' .. I suspect it will not be that easy though and is one of my longer term projects. Maybe its time to start ....

on (another) side note .. I have always found it surprising that high resolution images where captured before good quality audio. I now consider both photography and audio 2 dimensional (since you need to consider time as an important aspect) .. I guess it was lucky that chemistry is suited to capturing images before electronics came of age.

This is an issue I did a lot of reading about, a few years ago. I have a file of pdfs I downloaded at that time which I don't have url's for offhand.

The matter of consonants being important is exactly as you said. They're described in the literature as "cues". So "cue enhancement" is a good search term, as is "speech enhancement". Here's one by Sonmez et al on "consonant discrimination" technology - there's a free version of this paper available somewhere online.

On the more technical side, Kalman filters are a method of enhancement that's been studied recently. (I admit it, I have no idea what a Kalman filter is or how it works. but you might! Feel free to enlighten me.)

Ultimately, the specific frequencies involved in the sound of cue consonants are pretty well defined, so there's no reason they shouldn't be preferentially enhanced by a selective amplifier. But apparently in practice it's not quite that straightforward - or as you said, the hearing aid people would be doing a lot better for their thousands of dollars..

Speech is just vibrations of air molecules that transfer energy. Whats a vowel sound in your language may just be a mechanical vibration in a machine. My one tractor resonates at the letter 'Z' when loaded just right. Is it saying 'Z' to me or is it just a complex acoustical phenomenon?

Also whats a consonant sound like when spoken with an accent? Or which is what in a different language?

Speech to text software is presently the only real artificial method of converting sounds into words that can be understood that I am presently aware of.

If your hearing is that bad cochlear implants are likely the only way your going to ever hear anything properly again. But even then with some forms of nerve damage induced hearing losses even then its not going to help much or at all.

No pun intended, but, I'm all ears on this one. I did much unprotected shooting in my youth and am now paying for it. Haven't gone to an aid, yet, but am close.

You'd think that, after a while, practicality would rule out over vanity.

I have had good results with the aids from America Hears. These aids were so superior to two expensive different sets that I had before. I don't recall right now what they had to say ref what this post addresses but they are worth investigating. They were very detailed and courteous with my extensive inquiries.

I'm no expert on this, but I suspect that the root problem is that consonant sounds consist of very brief (though abrupt) transients. Vowel sounds are sustained longer (voiced), hence are easier to pick up and amplify.

Another problem is that a really high-fidelity amplifier may entail enough components that it is hard to squeeze them into a small space without sacrificing quality. Thus there must be compromises in the case of miniature devices. Al Mond's original post is excellent in the questions it poses and the ideas suggested. Same for several of the replies.

I found this pdf paper by Rodman, which gives an excellent overview of intelligibility issues re: consonants and fundamental frequencies. It's about telephone systems, but covers all the basics of intelligibility very coherently. The trick about consonants is that they contain most of the information in speech - he puts this very clearly.

Couldn't get the link to work, it's the pdf "Effect of Bandwidth on Speech Intelligibility" from polycom.com listed at the top here:

http://www.google.com/cse?q=bandwidth+intelligibility+Rodman+2006&sa=Search&cx=partner-pub-2070091971271392%3Aougxymc6y19&ie=UTF-8

What isn't discussed in this paper, is when there is a noise problem in the upper frequencies (think wind in a hearing aid, for example). I've used 9500 or 9000 hz as a cutoff frequency for speech intelligibility, to get rid of interference in the upper band - it seems to help. His data show that 7000 hz is a reasonable bandwidth for intelligibility.

There is other work that supports the view that the amplitude ratio between consonants and vowels is important. From the figures given by Rodman, vowels energy is generally below 3000 hz while consonants energy is generally above it. So a simple "intelligibility amplifier" might want to reduce energy below 3000 hz, raise energy from 3000 to 7000 hz, and maybe cut off above 8 or 9000 if there's interference in the high frequencies, for example.

It's also true that in hearing loss we may lose specific frequency bands - so custom adjustments would be a good idea. But anyone HOH can test out these simple ideas with a basic audio filtering program with lowpass and highpass filters. (I did my stuff with Audacity).

artsmith...sorry to say but nothing new if your post. If you just google 'audiogram' you will see that this is exactly the bands that are tested and hearing aid bandwidths are also in this range.

Hi,

Hearing and discriminating may involve correlation between the two ears.

This may suppress the noise from sources that are not in the same direction as the wanted source.

No human or mammal correlator in the brain is known (as far as I know, I am not an expert on hearing) but birds have it hard wired in a hollow bone that connects the two ears and is equipped with sensorial nerve cells along its length!

Birds are much better then we are, owls grasping mice below 5 to 10 cm of snow or in the dark.

As I have problems since the age of 17 when I did blow 1 cm3 of explosive unintended in 70 cm distance and had 3 days ringing in one ear and a drop in the 1500 to 2500 Hz band in that ear and a marked difficulty to hear and understand in noisy environments from then on - this is leading me to the conclusion that a correlator is existing and necessary inside our brain. A correlator that is tuned by our brain towards the direction where we want us to "focus" our hearing.

As we do not hear the phase of the many frequencies we can (or not) hear a new (not new in RADAR "hearing" may be not new in audio) approach might use the fact that the power spectrum is the Fourier transform of the autocorrelation function.

I would like also further developed the concept of direct coupling of sound to the bones of our skull - bypassing the impaired hearing system. Any microphone - amplifier - exciter will be possible. To be tuned in frequency response in the region of 30Hz to 10KHz (?) or 16 to 17 octaves each to be resolved by half tones. Will fit with considerable battery power into 2cm diameter and 5 to 8 mm height.

This may be much more powerful than any other approach - but two buttons behind/below the ears may be not accepted by everybody.

RHABE

I used to have a boss who wore glasses which had a built in hearing aid: they coupled the sound directly into the bone via the "behind the ear" section. They looked far more like conventional glasses than these.

bone conduction hearing system on spectacles

There is always constant research in this topic;

http://www.eurekalert.org/pub_releases/2009-08/afot-htw080509.php

Carefully glanced-thru existing posts, hoping this isn't redundant...

The TOPIC and opening line: "Has anyone tried to make an audio amplifier which produces more speech understandability" really rang loud-and-clear, but in the context of a different arena of application.

Deep sea commercial divers, when breathing heliox mixed gas, were unintelligible on the earliest / original intercomm units. Lots of science went into development of the helium unscrambler units that are used today.

I, myself (like another poster) suffered some hearing loss from hunting / target shooting and inadequately-protected 8" howitzer firing (observing). Even the current state-of-the-art unscramblers doesn't make the voices totally clear for me.

Sounds like an opportune time for SOMEONE to take the bull by the horns and organize whatever ongoing efforts exist to standardize the "voices" (and 'overall performance') in such tools as pocket translators, GPS units, etc.

IF things were, in fact standardized, transitioning from one country to another in the future (using the various "Help" units) might be facilitated greatly ... especially for us aging types...