salientAudio.py mthorogo at sfu dot ca last modified July 2013
implement pre-processing stage to normalize variables
Program to predict the valence and arousal of 4 second audio chunks. Algorithm pulled from http://nime2013.kaist.ac.kr/program/papers/day2/poster2/157/157_Paper.pdf
It uses a standard mlr algorithm trained with data derived from human subject study If your wanting to supply your own training set then pay attention to the feature extraction technique that can be gleened from the process method and the yaafeEngine. If you need assistance in creating such a training set you can ontact the author by email
Input: wav or aiff audio file >= 4seconds Output: csv file with valence, arousal, start time, end time
python salientAudio.py path/to/audio/file.wav(aif)
feature extraction http://yaafe.sourceforge.net/
OLS algorithm http://wiki.scipy.org/Cookbook/OLS
yaafeEngine.py and salientTraining.csv (should be with the repo)