Picking throat mic for use w/ EasyVR speech reco shield

Hope this is the right place, if not feel free to move the thread...

I am building a project using an EasyVR speech recognition shield https://www.sparkfun.com/products/10963? for control input.

I am likely to be using the item in a high-noise environment, so I would like to use a "throat mike" in place of the little microphone that comes with the kit.

According to the EasyVR manual, the microphone is fairly fussy, and they don't support anything other than the included mic, that is (IMHO) totally unsuited for the project. I found there are huge numbers of throat mikes on eBay (as just a first starting place to look) with most listings saying what radios they work with, but not necessarily what the actual specs on the mike are.

Per the Easy VR manual, the mike needs a 38dB sensitivity (0dB=1V/Pa @1KHz), 2.2Kohm load impedance, 3V operating voltage, and a nearly flat frequency response from 100Hz - 20KHz. I don't really need audio output, but most of the mikes seem to have a built in speaker, so the EasyVR output specs are either 8ohm speaker terminals or a 16-32ohm headphone jack.

Can anyone suggest a suitable microphone that would work? (cost is an object, so preferably one that is reasonably priced...)

Also, the manual says that the EasyVR folks have a forum of their own, but I can't find it on their website - anyone know where it is?

Thanks, ex-Gooserider


I'm a little lean on the electronics knowledge but I make my living with my recording studio by day. Speaking broadly, a mic is just a transducer that turns vibrations into AC voltage.

I'm not sure where to find mics of your exacting specs. I'm a little skeptical of a throat mic's ability to go up into the 20Khz ballpark...or ever 5Khz. This is not to say the mic itself won't pick it up, but I'm curious what kind of "information" (high frequency energy) can be found in the throat. It's my assumption this is almost entirely low frequency fundamental frequency stuff. How important that potentially missing high frequency energy is to your EasyVR is a total guess to me but the fact that they mention needing 100Hz-20Khz suggest high frequencies are important.

It is possible for a throat mic to compensate for the lack of high end, but I suspect it comes at the expense of isolation.

Then again, this could be a case of my background totally misleading me.

I found an audio clip of one (FINALLY!...Don't try Youtube.) http://www.iasus-concepts.com/nt/nt-3.htm I'm not hearing anything over 5k to speak of and could only guess what that means for speech recognition software, but the fact that a human has to squint a little bit to make out the words is a red flag.

How loud of environment are we talking? There may be other options other than the throat mic method.

Good luck!


I wondered about the clarity of throat mikes. I've heard them before, and found some difficulty understanding what was being said. A noise cancelling microphone might be a better option.

Something like http://www.amazon.com/Olympus-ME-52W-Noise-Canceling-Microphone/dp/B000MYPPPE possibly.

Possibly I could be using this in a VERY loud environment. I'm a paraplegic, and my application is essentially an electric foot pedal presser to enable me to use foot pedal equipped machines without needing to modify them.

The primary place I'm going to be using the device when I get it finished is a place in Somerville, MA called the Artisan's Asylum, which is one of the country's largest "hackerspaces". It includes some very loud machines that reach out even to the quieter spaces, but I'm likely to be using it in the welding shop on a TIG welder, with other people running angle grinders and the like just a few feet away...

The person who talked me into using the speech reco said that he used to make similar systems many years ago that used a throat mike to control pipewelding equipment in an oilfield environment, so presumably it can be done....

OTOH, the EasyVR FAQ said in one question that they didn't recommend a noise cancelling mike because it introduced to much distortion...

My primary concern is with meeting the electrical specs - I figure that even if the so called "speaker independent" vocab doesn't work right, I probably have more than enough speaker dependent slots available to train it for what I need - and if I use the same setup for training as I do for use, then any distortion won't matter....