1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
|
#!/usr/bin/env python
"""Implementation of entropy algorithm as described in:
**Krejtz K., Szmidt T., Duchowski A.T. (2014).**
*Entropy-based statistical analysis of eye movement transitions.*
Proceedings of the Symposium on Eye Tracking Research and Applications (ETRA'14, 159-166).
[https://doi.org/10.1145/2578153.2578176](https://doi.org/10.1145/2578153.2578176)
"""
__author__ = "Théo de la Hogue"
__credits__ = []
__copyright__ = "Copyright 2023, Ecole Nationale de l'Aviation Civile (ENAC)"
__license__ = "BSD"
from typing import Tuple
from dataclasses import dataclass
from argaze import GazeFeatures
import pandas
import numpy
@dataclass
class AOIScanPathAnalyzer(GazeFeatures.AOIScanPathAnalyzer):
def __post_init__(self):
pass
def analyze(self, aoi_scan_path: GazeFeatures.AOIScanPathType, transition_matrix_probabilities: pandas.DataFrame) -> Tuple[float, float]:
"""Analyze aoi scan path.
* **Returns:**
- stationary entropy
- transition entropy"""
assert(len(aoi_scan_path) > 1)
# Count total number of fixations and how many fixations are there per aoi
scan_fixations_count, aoi_fixations_count = aoi_scan_path.fixations_count()
# Probability to have a fixation onto each aoi
stationary_probalities = {aoi: count/scan_fixations_count for aoi, count in aoi_fixations_count.items()}
# Stationary entropy
stationary_entropy = 0
for aoi, p in stationary_probalities.items():
stationary_entropy += p * numpy.log(p + 1e-9)
stationary_entropy *= -1
# Transition entropy
transition_entropy = 0
destination_p_log_sum = transition_matrix_probabilities.apply(lambda row: row.apply(lambda p: p * numpy.log(p + 1e-9)).sum(), axis=1)
for aoi, s in destination_p_log_sum.items():
transition_entropy += s * stationary_probalities[aoi]
transition_entropy *= -1
return stationary_entropy, transition_entropy
|
