Grove Sound sensor - how to make it more precise?


#1

Hi,

I have a rapsbPi2 + GrovePi. I’m trying to make use of Grove sound sensor.
I’m trying to make it execute a command when I start speaking. I’m reading values every 200ms.

Even in a completely quiet room values it’s giving are in 240-370 range and when I start speaking really close to it it sometimes jumps to 400-450 and sometimes not. When I speak from more than 1 meter distance it doesn’t show any changes at all.

Should I try using approximation functions for last N values?
Maybe there’s a way to make it really sensitive to the frequencies of a human voice?

The values are just so random I’m not even sure how anyone is using it.

Are there any better solutions for recognizing when someone starts speaking?

Thanks


#2

Hi experigo,
We have found that the noise sensors is usually good enough to find the ambient noise level near it. It is not good for a lot of applications. Approximations over the last few value will definitely make the data better and it might be worth keeping the sound sensor inside a paper cone to make it a bit more sensitive. The sound sensor is designed that way, you can read about it in the comments section here: http://www.seeedstudio.com/depot/Grove-Sound-Sensor-p-752.html.

-Karan