Motion-Print Control Task Movements (38 subjects)

This dataset is composed of 4-Dimensional time series files, representing the movements of all 38 participants during a novel control task. In the ‘5D_Data_Extractor.py’ file this can be set up to 6-Dimension, by the ‘fields_included’ variable. Two folders are included, one ready for preprocessing (‘subjects raw’) and the other already preprocessed ‘subjects preprocessed’.

In the 'subjects raw' folder there is 1 file per subject, with all 15 runs of the task contained. Each run is separated by a line of all zeros. The raw data is sampled at 100 HZ, meaning 100 samples per second. Given this high sampling rate aggregation is done in preprocessing, to produce the files in the  'subjects preprocessed' folder. Depending on the ‘agg_rate’ given in the ‘5D_Data_Extractor.py’ file, each run at 100 HZ is sampled down by combining every ‘agg_rate’ data points (such as 100, 50 or 20) into one. 

In the 'subjects preprocessed' folder there are 2 files per subject, one 'TRAIN' and one 'TEST', which have been aggregated and split from the 'subjects raw' folder. These were made using ‘agg_rate’ of 100 (converting them to 1 HZ), ‘fields_included’ of ‘Ys’ and ‘Zs’, and ‘Dists’ and ‘test_runvals’ of 13, 14, 15. This means that of the 15 total runs, those three runs composed the TEST for each subject. The 'TRAIN' files are therefore larger, containing the subjects' movements on runs 1-12 of the task. The 'TRAIN' files were used to fit machine learning models for each subject, and the 'TEST' files were used to validate these models.