Digital Filter Controlled by Deep Learning Solution(camera tracking).
Following post can be somehow considered as a continuity of my previous article, where I described the design methodology of digital filter (main focus of notch filter, which removes certain frequency). However, this post does not extend given theory or performed simulations. Presented post focuses mainly on “final” project, and gives you great opportunity to be inspired to enjoy the technology (check my GitHub)
It this case I would like to display how you can control the pass band of the digital filer by hand (in this case RGB camera captures the position of you finger and moves in real time — in this project unfortunately with some latency, the pass band of the filter). In my case I used the notch filer so it will be straightforward for you to modify the project and apply the low pass filer or high pass filter.
Following project (as it was my intention) can be excellent opportunity to start your own application, where there is a need for such “touch-less” frequency domain adjustments. There are enormous area for discussed application, I think.
Below image indicates how the application works.
The position of your finger is detected by the RGB camera and deep neural network. For this task I applied state-of-the-art open source library — MediaPipe from Google (very easy installation).
pip3 install mediapipe
The position of the finger controls further the position of the notch filer in frequency domain, so it is easy to remove the frequency “on the fly”. On the screen you can see the spectrum of the signal and the position of the filter.
For this certain project I reused the design concept I displayed in my previous post.
Before I “connected filter” to camera I checked the repose — dynamic position changes of notch filter manually. I applied “slider” (see below) which can be used (instead of camera). By moving the slider you can see how the spectrum of notch filter is moving in frequency domain.
You can reuse concept and applied other type of the filter or control other parameter of you system. You can (for example) amplified signal, control equaliser, synthesizer, possibilities are enormous. All described concepts I discuss now you will find easy on my GitHub.
Final results!
Enjoy the project.
Thanks you for reading.
