Skip to content

A physics-informed neural network method for head-related transfer function upsampling

Notifications You must be signed in to change notification settings

feima1024/PINN-for-HRTF-upsampling

Repository files navigation

This is the source code for paper

"Spatial Upsampling of Head-Related Transfer Functions Using a Physics-Informed Neural Network" which is also attached here as a PDF file.

Sec. V, the interpolation experiment

Read pinn.py and see the result shown in 'interpolation.png' and 'interpolation.fig'


Or you can run the code and generate the result by yourself:

0, Download all files into one folder;

1, start a python terminal;

2, go to the same folder;

3, run the code by exec(open('pinn.py').read());

4, pinn.py will read the 40.mat file and generate the 40_L3.mat.

Trained one core of Macbook M1 pro, the runtime was about 7 hours.
The runtime can be significantly reduced by use more cores.

5, Start Matlab, and go to the same folder, run the 'result.m' file, then you will get the figures.


About

A physics-informed neural network method for head-related transfer function upsampling

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published