Main Site Documentation

Programmable sound sensor


#1

Hello All I am new to the forums, I also have very little computer and technical background(putting it out there). I am working on a project and I am wondering if there are sensors that can be programmed to recognize multiple sounds and specific sounds? Or can they be built and what kind of person could help me with something like that. I would like to do it all myself but with my limited education in this field I believe it may be tough. Any input would be greatly appreciated. I thank you in advance

Tom


#2

I’m not aware of any.

What types of sound are you interested in? You might be able to find a graphic equaliser chip that can detect certain frequencies.


#3

I am looking to make a sensor that can recognize fire/police/ambulance sirens.


#4

There are about 4 to 5 different siren tones


#5

Not that I’m an expert in that field, but I think a frequency analysis should do it.
The most common one is also known as Fourier analysis.

By this you can analyze the spectrum of an mix of frequencies like sound waves.
You get an graph with a peek at every frequency which is ‘strongest’ in the tone.
Each siren will/might have a different graph.
You might need to check if a specific sequence of graphs comes in a timely sequence.

If you connect this to a neuronal network, you could train the system to different sirens (or any other sound). But if you have clearly different sounds this should net be necessary.
That’s an complex topic on it’s own.

The best of all is that this is all achievable by software.
Not sure if the power of a NETMF board is sufficient, but I would think that at least the G400 or even the G120 could handle the frequency analysis when using RLP (native C code).
I think there is a RLP sample for audio recording, which would be a great starting point.


#6

Thanks for the info!!!


#7

You might be able to use something like the module in this video, i believe it uses MSGEQ7 chips to split up sound into 7 different frequency bands.

63Hz, 160Hz, 400Hz, 1kHz, 2.5kHz, 6.25kHz and 16kHz


#8

You would be best to start off with something like matlab or free alternative(octave) and run some fft s over some example recordings to see if you can see a simple pattern that distinguishes them. Work out what you need before you start building any hardware.

A lot of voice recognition uses wavelets rather than fft s so that’s an option too.

Alternately use Audacity and look at the spectagraph.


#9

Thank you all for the info. Most of this is above my education level. What type of person could be able to help me out with this? Electrical engineer? Computer engineer? Computer programmer? I will try and find someone local to physically help me here in central florida. Again thank you all for the info.


#10

It depends on how much time you have for this project.
With plenty of time anyone with a good technical understanding should be able to solve this problem. The community will help you if you need assistance on NETMF/RLP/Gadgeteer problems.

An electrical engineer with a bit programming experience should be able to do it.
A sound engineer should have the knowledge for the frequency stuff as well.
Any programmer/software engineer/… who reads into the topic of the frequency analysis should be able to solve it as well.

For the electrical part: If you use Gadgeteer components everything should be plug and play.

b.t.w. What kind of education/profession do you have? (If you don’t mind sharing this information)
How did you come to this project?


#11

A device that recognizes police sirens by tones… Not sure I understand the application of such a device… ???


#12

it’s for your car:
If you hear a sirene and green light goes on -> go on, it’s just an ambulance
red light: slow down, Police is coming :smiley:


#13

I have basic web design and a basic programming class I took about 8 years ago. The application will be for a gate(or a couple gates @ the entrance of a secure business. I would like them to open to the sirens.


#14

I know there is a device that does this but I can’t afforded to pay the $500 for it. Although I could probably sell it after


#15

If you need only one device, and assumed you need to pay for the hours and hardware invested into this project, I would pay the 500$.
I don’t think it will pay out for a single device. You also need to take in account that your device might not work at first (or eventually not at all).


#16

So you don’t need to identify what type of siron it is then. That makes it easier. I hear that some traffic lights detect the flashing lights of emergency vehicles and turn green.

Not very secure though.