This is a project I am doing in my free time at work to study how sonar and other sound localization technologies work. Currently, I can find out where a sound is originated only by azimuth.
I am working on creating a microphone array to achieve better precision and better estimate of true location. I am experimenting with recordings from an in line 4 mic array to start with, but I have a total of 8 microphone that I can use so I will eventually create a proper 2 dimensional array to give an attitude direction and a distance estimate.
To achieve this, I am using beamforming algorithms which works by shifting the microphone recordings by a certain delay until the sound power reaches its peak. This works because each microphone is spaced away from eachother by the same distance, so a sound will reach each microphones at different intervals. By looking at which microphone were delayed and by how much, we can then get an estimate of the location of the sound.
I am making use of the Pyroomacoustics library in this project as it does a lot of heavy lifting with the complex algorithms used and their great simulation functionality. This being just a proof of concept/prototype, I am not worried about writting everything from scratch, but rather learn the audio concepts and how to apply them in code.
- Find azimuth from stereo microphone recording using GCC PHAT
- Simulate a 4 microphone line array and compare which algorithm is best
- Find azimuth of a static sound from a real 4 microphone line array
- Modify the program to handle moving sounds
- Build a 2 dimenional mic array
- Find azimuth and attitude of sound and diplay it nicely