StructureFromSound

Code for the Structure from Sound project (SFS) The aim of the project is to develop tools for estimating microphone positions and sound source movements using a set of audio files as input only.

This is a challenging parameter estimation problem. For the project we also develop tools that can be of use for other TDOA type problems.

If you use this code, please cite

Simayijiang Zhayida, Simon Segerblom Rex, Yubin Kuang, Fredrik Andersson, Kalle {\AA}str{"o}m}, An Automatic System for Acoustic Microphone Geometry Calibration based on Minimal Solvers, [https://arxiv.org/abs/1610.02392](arXiv preprint arXiv:1610.02392), 2016.

@article{simayijiang2016automatic, title={An Automatic System for Acoustic Microphone Geometry Calibration based on Minimal Solvers}, author={Simayijiang Zhayida, Simon Segerblom Rex, Yubin Kuang, Fredrik Andersson, Kalle {\AA}str{"o}m}, journal={arXiv preprint arXiv:1610.02392}, year={2016} }

Other relevant papers are:

@inproceedings{simayijiang2014automatic, title={An Automatic System for Microphone Self-Localization Using Ambient Sound}, author={Simayijiang, Zhayida and Andersson, Fredrik and Kuang, Yubin and {\AA}str{"o}m, Kalle}, booktitle={European Signal Processing Conference (Eusipco 2014)}, pages={5}, year={2014}, organization={EURASIP (European Association for Signal Processing)} }

@inbook{393831000cb64d7988c470971301d957, title = "Stratified Sensor Network Self-Calibration From TDOA Measurements", keywords = "Network Self-calibration, TDOA, TOA, Minimal Problem", author = "Yubin Kuang and Kalle {\AA}str{"o}m", year = "2013", booktitle = "[European Signal Processing Conference (Eusipco 2014)]", publisher = "IEEE--Institute of Electrical and Electronics Engineers Inc.", }

Code

Run main.m

Data

Example data for the system can be downloaded from the StructureFromSoundDatabase (sfsdb) at http://vision.maths.lth.se/sfsdb/

Briefly the datasets are:

1. bassh1 - lunch room (3DR, 3DS, Cont, Echo, Single, Distinct)
2. bassh2 - lunch room (3DR, 3DS, Cont, Echo, Single, Distinct)
3. bassh3 - lunch room (3DR, 3DS, Cont, Echo, Single, Distinct)
4. grieg1 - lecture room (3DR, 3DS, Cont, Echo, Single, Distinct)
5. grieg2 - lecture room (3DR, 3DS, Cont, Echo, Single, Distinct)
6. grieg3 - lecture room (3DR, 3DS, Cont, Echo, Single, Distinct)
7. grieg4 - lecture room (3DR, 3DS, Cont, Echo, Single, Distinct)
8. grieg5 - lecture room (3DR, 3DS, Cont, Echo, Single, Distinct)
9. spacecure1 - anechoic chamber (3DR, 3DS, Cont, NoEcho, Single, Distinct)
10. whywereyouawayayearroy - anechoic chamber (3DR, 3DS, Cont, NoEcho, Multiple, NotDistinct)
11. axelf - kitchen (2DR, 3DS, Cont, Echo, Single, Distinct)
12. gone - kitchen (2DR, 3DS, Cont, Echo, Single, Distinct)

Here

• 3DR - Dimension of affine hull of receivers is 3. (Microphones not in a plane)
• 2DR - Dimension of affine hull of receivers is 2. (Microphones in a plane)
• 1DR - Dimension of affine hull of receivers is 1. (Microphones on a line)
• (0DR - Dimension of affine hull of receivers is 0. (Microphones all in the same position)
• 3DS- Dimension of affine hull of sound events is 3. (Sound events not in a plane)
• 2DS- Dimension of affine hull of sound events is 2. (Sound events in a plane)
• 1DS - Dimension of affine hull of sound events is 1. (Sound events on a line)
• (0DS Dimension of affine hull of sound events is 0. (Sound events all in the same position)
• Claps - Claps, a few distinct sound events
• Cont - Continuous sound events
• Echo - In a normal (reverberant) room
• NoEcho - In anaechoic chamber
• Single - Single sound source
• Multiple - Multiple sound sources
• Distinct - Spatially distinct sound sources
• NotDistinct - Spatially spread out sound sources

Experiments in lunch room September 18, 2014

The microphone setup is illustrated in the following figure.

We made 3 experiments (bassh1, bassh2, bassh3). There are several still images of the microphone setup and one stationary few film recording of the experiments.

Experiments in lecture room on November 27, 2014

The microphone setup is illustrated in the following figure.

We made 5 experiments (greig1 to grieg5). There are several still images of the microphone setup and a few film recordings of the experiment. During theses film recordings, the camera was moving.