Accelerometer and annotation data for trained assistance dogs performing signalling behaviour on instruction

11 trained dogs each wore an AX3 accelerometer attached to their neck collar, for periods of 10 days on average, sampling XYZ direction at 12.5Hz. During this period they  performed a variety of "tricks" on instruction from their owner, such as sit, begup, spin, etc. These were recorded on timestamped videos which were then annotated manually to mark the date/time for each of these behaviours-on-demand and this is the annotations file.  These can act as a ground truth for training a machine learning algorithm to recognise canine behaviours from the accelerometer data.

11 CSV files, one for each dog have 4 fields for each row corresponding to the date and time, x, y, and z directional magnitude values, sampled at 12.5 Hz, so 60 x 12.5 = 750 values per minute.

The annotation file has 9 fields per row corresponding to a record identifier, the x, y and z directional magnitude values, the sum or Amag value, the 4 character code for the participating dog which matches with the name of a CSV file, the trick or behaviour such as sit, spin anti-clockwise, stand, etc., the time and date of the activity which will match the  timestamps in the 11 CSV files.

The Mapping Participant to paper index Excel sheet is the mapping of which dog (4 character code) refers to which ID in the paper published about this work.