Do microphones deteriorate with time?

[chrisbainbridge]

I use a headband mounted boom microphone for voice recognition to dictate large reports. I have used the same microphone for 8+ years and always been very happy.  Recently there has been a small but noticeable decrease in recognition accuracy.

Could this be due to the microphone wearing out, being covered in saliva droplets, etc or is it more likely to be the software picking up small incremental glitches from when i do not fully correct it?

Thank you

[Kim]

Record some audio through the microphone with Audacity or something and listen to it.  If it sounds crap, particularly as a function of cable-wiggling, there's your problem.

[chrisbainbridge]

Hi Kim

The microphone is bluetooth and then USB cable direct to the computer.  If the microphone has got to that level of damage then the voice recognition will be zero.  We are talking about a less than 5% deterioration in voice recognition ability.  Think but rally annoying in terms of time wasted doing the corrections.

[Kim]

Yeah, but you're suggesting using voice recognition software to diagnose a suspected problem with audio hardware.  That's daft.  Listen to the audio.  If it's intermittently cutting out, or crackling in response to plosives, or slightly distorted, or full of 50Hz mains hum or whatever, that'll tell you it's a microphone problem.  If the audio sounds good, then you know it's something else affecting the voice recognition performance and you can look elsewhere.

[chrisbainbridge]

When I am looking for 98% or better accuracy even having the boom microphone out of position by 1cm will drop accuracy down to unintelligible.  The sound profile has built up over about 8 years and does accrete a series of small mistakes such that every few years or when a new version comes out I buy a new computer and reinstall the whole thing.  The only thing i have never changed is the microphone and i am simply asking if the response of the microphone changes with age and all the aerosolised gubbins coming out of my mouth even at an angle.  I know that everything is working to better than 90% + and I know that I will not be able to hear a problem with the audio by ear.

[Kim]

Well in that case, the answer is "Yes, it might."  I expect that all sorts of things could happen to affect the performance.  Anything from electrolytic capacitors drying out to dodgy lead-free solder joints to gunk clogging up the orifice to glue coming unstuck or loose self-untapping screws changing the acoustics of the enclosure might have a subtle effect on the performance in ways that might or might not be apparent by ear or spectrogram.

Or it might be that the algorithm's been updated, it's trained itself on bad data, or some other factor is reducing the computing performance available for speech recognition.  Without doing systematic diagnostics, who knows?

[slowfen]

As an aside

I used via voice for a number of years, but found that half a bottle of wine would decrease the accuracy of the recognition ( I even toyed with the idea of setting up a second account for when drinking)

Query has there been a change in your voice??

[chrisbainbridge]

As an aside

I used via voice for a number of years, but found that half a bottle of wine would decrease the accuracy of the recognition ( I even toyed with the idea of setting up a second account for when drinking)

Query has there been a change in your voice??

good suggestion.  i know that my accuracy changes even when i am starting to get a cold.  i could almost use it as a diagnostic test!  I am 4-5 years older than when I established this original profile.  i normally change computer and profile with each new version of naturally Speaking but missed the last cycle.

Thank you for the suggestion.  I think it reinforces my plan for new microphone and new computer.

[Valiant]

Microphones can and do deteriorate over time as the suspension etc gets looser but it does take a fair bit of time and loads of bashing about.