EyeSwipe2 Data Analysis

This is an initial analysis of the data collected in the experiment comparing EyeSwipe1 to EyeSwipe2. The 13 participants performed 16 (distributed equally in 4 different days) 5-minute sessions with each keyboard. After each typing session the participants performed a layout expertise session with half of the keys in the keyboard. The two layout modes (hidden or visible) alternated after each typing session.

Some participants had problems with the calibration in all sessions and will be removed from the analysis. Other participants had calibration problems in specific days or sessions. We still have to decide how to handle these cases.

Interacting with the visualizations

The plots are interactive, but GitHub can't display the interactive widgets. To be able to interact with the data you will need to clone this repository and run jupyter notebook from the root folder.

You can use the controls to hide/show data from a particular participant.



In [1]:

    
%matplotlib notebook
# Imports
import pandas as pd
from data_handler import load_data, compute_mean_values, DataPlotter

typing_df, gesture_df, layout_df, participants = load_data()

# Remove participants with severe calibration problems
participants.remove('A04')
participants.remove('A08')
participants.remove('A09')
participants.remove('A11')
typing_df = typing_df[typing_df['participant_id'].isin(participants)]
gesture_df = gesture_df[gesture_df['participant_id'].isin(participants)]
layout_df = layout_df[layout_df['participant_id'].isin(participants)]

Sentence count

The following plot shows the number sentences typed by each participant.



In [12]:

    
sentence_count = typing_df.groupby(('participant_id', 'mode_id', 'session_id'))['trial_id'].count().reset_index()
sentence_count.rename(columns={'trial_id': 'trials'}, inplace=True)
DataPlotter('trials', sentence_count, participants).show()

Words Per Minute (wpm)

The word per minute (wpm) rate is shown in the figure below.



In [2]:

    
DataPlotter('wpm', typing_df, participants).show(ylim=(0, 25))

Mean wpm per session



In [3]:

    
wpm1 = compute_mean_values(typing_df, 'eyeswipe1', 'wpm')
wpm1.columns = ['EyeSwipe1 (wpm)']
wpm2 = compute_mean_values(typing_df, 'eyeswipe2', 'wpm')
wpm2.columns = ['EyeSwipe2 (wpm)']
pd.concat([wpm1, wpm2], axis=1, join='inner')









    Out[3]:






  
    
      
      EyeSwipe1 (wpm)
      EyeSwipe2 (wpm)
    
    
      session_id
      
      
    
  
  
    
      1
      7.576323
      10.762957
    
    
      2
      9.007117
      12.007915
    
    
      3
      9.794918
      11.968313
    
    
      4
      10.378821
      13.586039
    
    
      5
      10.248143
      14.335905
    
    
      6
      11.145737
      15.179085
    
    
      7
      11.122374
      15.369142
    
    
      8
      12.007789
      14.984717
    
    
      9
      10.896243
      14.675383
    
    
      10
      11.775933
      15.285568
    
    
      11
      12.202381
      14.811000
    
    
      12
      12.946275
      15.147822
    
    
      13
      12.228889
      14.986438
    
    
      14
      12.208756
      15.996097
    
    
      15
      12.630720
      16.052001
    
    
      16
      12.835639
      15.344447

Error rate

The Minimum String Distance (MSD) error rate is computed using the Levenshtein distance (MSD) between the typed and expected sentences. It is given by:

\begin{equation} ErrorRate = \frac{MSD(typed, expected)}{\max(|typed|, |expected|)} \end{equation}



In [4]:

    
# Compute length of typed and expected sentences
typing_df['lexp'] = typing_df['expected_sentence'].str.len()
typing_df['ltyp'] = typing_df['typed_sentence'].str.len()
# Compute MSD error rate
typing_df['error rate'] = typing_df.msd / typing_df[['lexp', 'ltyp']].max(axis=1) * 100
# Plot
DataPlotter('error rate', typing_df, participants).show(ylim=(0, 20))

Selection count

The number of selections before a key was chosen is shown in the next figure. This information should be considered in combination with the following plot, because the participant might have made few selections, but chosen the wrong letter.



In [5]:

    
DataPlotter('selection_count', layout_df, participants).show(ylim=(0, 6))

Distance to target

The distance (in pixels) between the center of the selected key to the center of the target key is shown in the plot below. The key width/height was about 100 px. The error was clearly larger for the hidden condition, as expected.



In [6]:

    
DataPlotter('distance_to_target', layout_df, participants).show()



In [7]:

    
DataPlotter('reaction_time', layout_df, participants).show(ylim=(0, 800))

Some considerations

Some participants had to recalibrate between the typing and layout sessions or even during the layout sessions. For this reason we may not be able to use their data to evaluate the dynamic calibration.



In [ ]:

	EyeSwipe1 (wpm)	EyeSwipe2 (wpm)
session_id
1	7.576323	10.762957
2	9.007117	12.007915
3	9.794918	11.968313
4	10.378821	13.586039
5	10.248143	14.335905
6	11.145737	15.179085
7	11.122374	15.369142
8	12.007789	14.984717
9	10.896243	14.675383
10	11.775933	15.285568
11	12.202381	14.811000
12	12.946275	15.147822
13	12.228889	14.986438
14	12.208756	15.996097
15	12.630720	16.052001
16	12.835639	15.344447