Mercurial > hg > dcase2013_sc_rnh
annotate README.md @ 2:def2b3fa1450 tip master
corrected README
| author | Gerard Roma <gerard.roma@upf.edu> |
|---|---|
| date | Mon, 04 Nov 2013 10:46:05 +0000 |
| parents | 96b1b8697b60 |
| children |
| rev | line source |
|---|---|
| gerard@2 | 1 The files in this repository represent our submission for the Scene Classificatoin task of the IEEE D-CASE AASP Challenge (http://c4dm.eecs.qmul.ac.uk/sceneseventschallenge/) |
| gerard@1 | 2 |
| gerard@2 | 3 The code has been tested mainly on Matlab2012 on OSX. |
| gerard@2 | 4 |
| gerard@1 | 5 Required libraries: rastamat and libsvm |
| gerard@1 | 6 |
| gerard@1 | 7 The implemented approach is described in: |
| gerard@2 | 8 |
| gerard@1 | 9 G. Roma, W. Nogueira, P.Herrera, _Recurrence Quantification Analysis Features for Environmental Sound Recognition_. Proceedings of IEEE Workshop on Applications of Signal Processing to Audio and Acoustics. New Paltz, USA 2013. |
| gerard@1 | 10 |
| gerard@1 | 11 The main files are: |
| gerard@1 | 12 |
| gerard@1 | 13 * classify_scenes.m |
| gerard@1 | 14 * analyze_files.m |
| gerard@1 | 15 * RQA.m |
| gerard@1 | 16 |
| gerard@2 | 17 The rest of matlab files can be used to test the code. Some of the files are taken from https://soundsoftware.ac.uk/projects/aasp-d-case-metrics |
| gerard@1 | 18 |
| gerard@1 | 19 Two submissions were sent to the challenge, one uses hardcoded SVM parameters, the other does grid search: |
| gerard@1 | 20 |
| gerard@1 | 21 classify_scenes(tmp_path, train_file,test_file, output_file, 0) % use hardcoded parameters for SVM |
| gerard@1 | 22 classify_scenes(tmp_path, train_file,test_file, output_file, 1) % use grid search |
| gerard@1 | 23 |
| gerard@2 | 24 temp_path is used to store features. |
| gerard@1 | 25 |