Human label file formatΒΆ

The test results are called human labels in the following and the average results of the single experiments are all stored in the same way in the human label file format. This is a simple csv file, which includes the following entries:

# Description of the results stored in the file
# stimuli, rating, 95% confidence interval
experiments/link_to/brs_file1.wav, -0.4653, 0.0123
experiments/link_to/brs_file2.wav,  0.2738, 0.1548

The file starts with a header that uses # as a comment sign and then includes at least three columns. The first one provides a link to the actual BRS file used in the experiment. The second one the average result from the experiment, this could be a mean, a median, or something else, and the third one showing the variance of the data, in most of the cases in the form of the confidence interval. The files could of course have more columns with additional information, like the time the listeners needed to response or if another value like the sound pressure level was changed during the experiment it could be indicated in a later column.

Besides the human label files, most experiments provide the anonymised results from single listeners. The format of those data can vary, but in all cases they are provided as plain text or csv files.