Looking for microphone suggestions, recording speach in crowded place - restaurant

[KKND]

Hey,

i'm looking for microphone that i can built in in a device with speach detection module.
Tried few already but resoults are not good enough. I'm having trouble with cutting background sound so the speach is well audible

Environment is quite difficult - lets say its a restaurant like Mcdonald with many customers arround.

Device is runing standard win10 with USB ports available. Customer is standing directly opposite the device.

Regards !

[Tim Gillett]

Welcome. Usually in a noisy place there's no substitute for getting the mic close to the person(s) of interest. For example singers on a noisy stage usually use mics only a few inches from their mouths. Mobile phone conversations can often be clearly understood even in fairly noisy places because the phone's mic is close to the person's mouth.

Hope this helps.

[KKND]

Yeah, problem is that customer stands about 40-70 cm away from mic.

[Sam Spoons]

What is the purpose of the recording? And what mic are you using now?

[KKND]

I'am building voice controlled device.

For tests im using Philips LFH3500/00 but it has to be built in mic or some kind of box module attached to main device.

[zenguitar]

The problem here is the distance from the microphone.

There are different designs of microphone polar patterns which are more sensitive to sound coming from the front and less sensitive to unwanted sounds from the side. You can read more in this article from a microphone manufacturer.

http://www.rode.com/blog/all/what-are-a ... r-patterns

For stage singers who are close to a microphone, this is great. But the further away you get from the microphone, the more critical it is that the microphone is aimed accurately at the singer/speaker. For your intended design in a restaurant the people speaking will not understand the need to position themselves correctly.

Of course, a well chosen good quality microphone will provide the best possible input. But I believe the solution to your problem is in designing software that can remove the unwanted background noise to highlight the wanted sound.

Rather than look at audio recording tools, you might want to look at the technology used by devices like Alexa. These home devices are designed to listen to a microphone and then apply specialist algorithms to identify what the speaker is saying.

Andy

[CS70]

Not sure you can solve it by microphone.. my first idea if I were to develop such a device would be to employ spectral analysis and machine learning to try and separate voices and then pick up the loudest one, according to some definition whose parameters would have to be learned in the field. Bit what Siri&c do.

Using two mics in a close stereo configuration would allow you to triangulate.

Even if the tech (and the science) mostly exists, It's not trivial, and far from immune to errors, but it's pretty much the same job our brain does biologically using our omni mics - the ears.

A sizable amount of R&D would be needed but a patented result would be very valuable and worth it..

[Sam Spoons]

Cool, Alexa manages it reasonable well for a limited number of commands, not sure how it'd work in a noisy environment but I suspect very clever processing is the key. For your situation, if you can confine the customer to a small area when they are speaking, a directional mic combined with some sort of real time noise cancelling (a second mic far enough from the customer to pick up the same background as the main mic but little of the voice) would be my best suggestion. That would reduce the load on the VR software to some extent.

[Tim Gillett]

The LFH3500 seems to have a uni directional mic (good). But the photo in the brochure shows it used handheld and close to the mouth in a dictation situation. Used at a distance of 400mm to 700mm AND in a noisy place I am not surprised it would struggle.