Skip to content

Latest commit

 

History

History
 
 

supplementary_material

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
assets
assets
 
 
README.md
README.md
 
 

README.md

Supplementary Material

This directory contains the supplementary material of the paper "Object size prediction from hand movement using a single RGB sensor". The following information is presented below:

  • the confusion matrices for each movement completion interval (20%, 40%, 60%, 80%, 100%), feature set (Ws-D, Ws-I) and data-split strategy (all-in, one-out).
  • the feature importance heatmaps for each movement completion interval (20%, 40%, 60%, 80%, 100%) and feature set (Ws-D, Ws-I), as computed for the Extra Trees model and the "all-in" data split.
  • the progression of each kinematic variable (TI-Ap, TM-Ap, IM-Ap, Wrist x-coordinate, Wrist y-coordinate) over the R-t-G movement.
  • a table containing the average and the standard deviation of the accuracy rates for each movement completion interval (20%, 40%, 60%, 80%, 100%), feature set (Ws-D, Ws-I) and data-split strategy (all-in, one-out).
  • the accuracy boxlots for each movement completion interval (20%, 40%, 60%, 80%, 100%), feature set (Ws-D, Ws-I) and data-split strategy (all-in, one-out).

Confusion Matrices

"all-in" data split
Ws-D feature set


"all-in" data split
Ws-I feature set


"one-out" data split
Ws-D feature set


"one-out" data split
Ws-I feature set

Feature Importances

Ws-D feature set


Ws-I feature set

Kinematic Variables

Finger Apertures (TI-Ap, TM-Ap, IM-Ap) Kinematic Variables


Wrist x-/y- Pixel Coordinates Kinematic Variables

Results

The average and the standard deviation (in parentheses) of the accuracy rates for each of the movement completion intervals (20%, 40%, 60%, 80%, 100%), data split strategies (all-in, one-out) and feature sets (Ws-D, Ws-I), as presented in Section 4 of the paper.

Data Split Workspace Dependency 20% 40% 60% 80% 100%
all-in Ws-D 68.4 (3.9) 81.5 (2.8) 87.3 (3.1) 91.3 (3.8) 94.3 (2.1)
all-in Ws-I 60.7 (5.7) 77.5 (4.3) 81.8 (4.0) 87.7 (3.6) 90.8 (2.9)
one-out Ws-D 44.0 (12.9) 55.1 (14.6) 56.5 (14.4) 62.2 (15.4) 67.1 (13.7)
one-out Ws-I 48.9 (12.1) 61.3 (8.9) 66.6 (9.2) 63.8 (11.5) 67.1 (15.0)

The corresponding boxplot figures are the following:

    "all-in" data split


    "one-out" data split