0
\$\begingroup\$

I'm taking a digital signal processing course and I have a project to create a data set with a microcontroller and process it for this course. I want to count only dog bark for example not dog growl. I have to use fft or similar algorithm for my project. I think I can use fft algorithm for counting dog bark time but as you can see I have never record data set and process it. I need some advice and I have some question. I am thinking of use RP2040 , SD card for data record and SPW2430 . I will use nonblocking sd card example.

I will make little pcb for this ( Copper Plate Pcb not proffesional ) and I willa ttach this PCB to dogs collar.

  1. I used to do a project with fft. Sensitive to certain frequencies of sound. Can I realize this project if I record the dog barking as analog and count it by looking at the frequency with fft?
  2. I am thinking of use SPW2430. No additional opamp is included and the output peak-to-peak voltage has a 0.67V DC bias. I will put this pcb to dog's collar. Do you think this microphone enough for record or should I use op-amp ?
  3. Which sampling rate should I pick for recording sounds in analog ? at least for this project (tune detect).

Maybe I have other problems that I don't know about, I think you know these issues much better.

\$\endgroup\$
2
  • \$\begingroup\$ If you want a reliable system, you need to look at the bark envelope as well as the frequency content. Think of a bark as a musical instrument: en.wikipedia.org/wiki/Envelope_(music) Load some pre-recorded barks into an app such as Audacity to see what I mean. My baseline would be: sample at 40 kHz, 1024 samples, therefore data is processed at 40kHz/1024, or about 39 Hz. \$\endgroup\$
    – Mattman944
    Commented Feb 19, 2023 at 15:15
  • \$\begingroup\$ @Mattman944 thank you. \$\endgroup\$
    – karl0101
    Commented Feb 20, 2023 at 4:31

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.