https://github.com/trishume/popclick
Detecting lip popping noises to trigger an action.
https://github.com/trishume/popclick
algorithm audio-recognition spectrograph
Last synced: 4 months ago
JSON representation
Detecting lip popping noises to trigger an action.
- Host: GitHub
- URL: https://github.com/trishume/popclick
- Owner: trishume
- Created: 2014-11-26T02:36:10.000Z (about 11 years ago)
- Default Branch: master
- Last Pushed: 2016-03-17T22:08:10.000Z (over 9 years ago)
- Last Synced: 2025-04-10T04:49:56.909Z (8 months ago)
- Topics: algorithm, audio-recognition, spectrograph
- Language: C++
- Size: 46.9 KB
- Stars: 16
- Watchers: 3
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: Readme.md
Awesome Lists containing this project
README
# Lip Popping Recognizer
This project implements a simple algorithm for recognizing when the user
makes a popping noise with their lips. The plan is to use this in concert
with my [eye tracker](https://theeyetribe.com/) as a way to click without
using my hands.
It is currently being rewritten as a [Vamp plugin](http://vamp-plugins.org/) in C++.
I develop the Vamp plugin in Sublime Text, compile it to a dylib and then load it in [Sonic Visualizer](http://www.sonicvisualiser.org/)
which is a great tool for developing audio recognition algorithms.
I plan on expanding this to recognize other mouth noises. This is an unsolved
problem in audio recognition as it has very different goals from other domains like speech recognition:
1. Noises are much easier to distinguish from each other. You can look at a spectrograph and identify if something is the noise you want, not so for a spoken word.
2. Goal is to be resource efficient and real-time. Generally speech recognizers don't do this because good results require heavy processing.
3. Highly reliable, very low false positives only a few false negatives. This is only possible because of the ease of distinguishing.
## Sonic Visualizer Screenshot
The top row is a spectrograph, the middle is a debug visualization of the algorithm state, the bottom row is the waveform and some parameters.
The red lines on the bottom row are when a lip pop can be recognized (if it was real time), the three lip pops and three non-pops are correctly classified.
This is the layout I use for debugging the algorithm, [Sonic Visualizer](http://www.sonicvisualiser.org/) allows me to inspect values and scroll and zoom around the test audio file.
## Building
Run `make -f Makefile.osx` on osx, change the `.osx` to the correct Makefile if you are on another platform.