I am working on a personal design project. i have attached a high level block diagram and generic parts list.

The device will function as an audio click detector and click amplifier. The device must be able to detect low amplitude "clicks" buried in other background noise. Picture a metronome clicking while a radio is playing in the same room. the device must be able to detect the clicks, ignore the background noise and amplify the clicks only.

i would like some suggestions on circuit design. I.e. should the circuit be analog with band pass filters or will the circuit have to do an analog to digital conversion and do a spectrum analysis of the audio waveform to be able to pick out the recurring pulse stream?

i would also like some suggestions on the component part numbers so i can
start to breadboard the circuit and troubleshoot it.

i'm sorry this turned into such a long post. i appreciate any questions and help.
tenfour

 

A circuit to accomplish this would be enormously difficult compared to a software solution on a computer. Do you have size, cost, power constraints, etc. that prevent using a computer?

Is the clicking at a predictable frequency? It gets a bit easier if you know to look for and lock onto, say, 80bpm.

 

wayneh,
yes i have size, cost and power constraints that would prevent a final package that would include using a computer as part of the final design. however, i could use a computer in the prototype design and then put any software code on an embedded microprocessor.

yes the audio clicking is at a predictable frequency and your guess is very close, 80bpm. please respond back if this information helps and thanks for your input so far.
tenfour

 

wayneh,
i did some more testing today. i recorded the clicks ran the fille against a spectrum analyzer. it looks like the clicks are about 2 beats/second and pretty low amplitude.
i did a second recording with bob dylan's "times are a changing" playing in the background and ran the second file against the fast fourier xform spectrum analyzer and the clicks were hidden by the audio in the background.

 

Getting a computer/analogue circuit to do what the brain finds simple can be extremely difficult, if not impossible. Just look at the fields of OCR (optical character recognition) or speech recognition - the first thing that gets done to the incoming signal is de-noising, usually by means of making the signal to be analysed by far and away the most dominant component.

Maybe some kind of noise cancelling equipment (as found in some stereo headphones) might help?

Alternatively, could the problem be approached from a different perspective? What is it that you are actually trying to achieve?

 

P-MONKE,
thx for the reply. yes the human brain is amazing and this device is not needed for most people who don't have a hearing problem. the application for this device would be for people who are hard of hearing i.e. "ME". it is a safety
application. if you look at my original blk diagram you can see that i want to
detect the clicks and if they don't go away in about 5 seconds then amplify the clicks to warn the hard of hearing person that the clicks are still present.

regarding your comment on speech recognition, i think the device will have to
"learn" the click pattern. i.e. amplitude, frequency, pulse with, and pulse interval so the spectrum analyzer can find the pulse stream with audio in the background.

there are alternative methods of solving the problem by hardwiring the device and using electrical signals from the clicks but i want to have the device to be more universal --- standalone not hardwired.
tenfour

 

It is a safety application.

Although I'm intrigued with your project, as you state it's a safety application, it may fall outside the "comfort zone" for this site.

I only say this as automotive applications are not allowed to be discussed due to concerns that any subsequent related incident may result in litigation. This is sad, but an accurate reflection of the mentality of a small proportion of the population who won't take responsibility for their own (usually stupid) actions.

I think the moderators should decide this one.