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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
| import numpy as np
from typing import Tuple
def detect_steering_reversals(
steering_angle: np.ndarray,
fps: int = 30,
min_angle_change: float = 3.0,
max_duration: float = 0.5,
) -> Tuple[np.ndarray, float]:
"""
检测转向盘微修正（reversals）
基于Toyota IDD数据集发现的酒精特异性特征：
酒驾驾驶员会出现高频微小转向修正
Args:
steering_angle: 转向角序列 (degrees), shape=(N,)
fps: 采样率
min_angle_change: 最小角度变化阈值
max_duration: 持续时间上限
Returns:
reversal_indices: 微修正位置索引
reversal_rate: 每分钟微修正次数
Example:
>>> # 模拟酒驾数据（高频微修正）
>>> np.random.seed(42)
>>> angles = np.cumsum(np.random.normal(0, 0.5, 1800))
>>> # 加入微修正
>>> for i in range(0, 1800, 50):
... angles[i:i+5] += np.random.normal(0, 3, 5)
>>> idx, rate = detect_steering_reversals(angles)
>>> print(f"Reversal rate: {rate:.1f}/min")
Reversal rate: 18.2/min
"""
diff = np.diff(steering_angle)
sign_changes = np.where(np.diff(np.sign(diff)) != 0)[0]
valid_reversals = []
for idx in sign_changes:
window_size = int(max_duration * fps)
start = max(0, idx - window_size)
end = min(len(steering_angle), idx + window_size)
angle_range = np.max(steering_angle[start:end]) - np.min(steering_angle[start:end])
if angle_range >= min_angle_change:
valid_reversals.append(idx)
valid_reversals = np.array(valid_reversals)
duration_min = len(steering_angle) / (fps * 60)
reversal_rate = len(valid_reversals) / duration_min if duration_min > 0 else 0
return valid_reversals, reversal_rate
def classify_impairment(
steering_data: np.ndarray,
gaze_entropy: float,
pupil_diameter_var: float,
speed_variance: float,
) -> dict:
"""
基于IDD数据集发现的多模态损伤分类
使用掩蔽效应感知的融合策略：
- 不依赖单一指标（避免掩蔽效应）
- 转向微修正 → 酒驾特异
- 注视熵下降 → 通用损伤
- 瞳孔直径变异 → 认知负荷
- 速度方差 → 辅助（不可靠）
Args:
steering_data: 转向角序列
gaze_entropy: 注视分布熵 (0-1)
pupil_diameter_var: 瞳孔直径方差
speed_variance: 速度方差
Returns:
classification: {
'state': 'normal|alcohol|cognitive|combined',
'confidence': float,
'evidence': dict
}
"""
_, reversal_rate = detect_steering_reversals(steering_data)
alcohol_score = 0.0
cognitive_score = 0.0
if reversal_rate > 15.0:
alcohol_score += 0.4
if gaze_entropy < 0.6:
alcohol_score += 0.2
cognitive_score += 0.3
if pupil_diameter_var > 0.15:
cognitive_score += 0.3
if speed_variance < 2.0:
cognitive_score += 0.1
elif speed_variance > 8.0:
alcohol_score += 0.1
if alcohol_score > 0.5 and cognitive_score > 0.4:
state = 'combined'
confidence = min(alcohol_score + cognitive_score, 0.95)
elif alcohol_score > cognitive_score:
state = 'alcohol'
confidence = alcohol_score
else:
state = 'cognitive'
confidence = cognitive_score
return {
'state': state,
'confidence': confidence,
'evidence': {
'reversal_rate': reversal_rate,
'gaze_entropy': gaze_entropy,
'pupil_var': pupil_diameter_var,
'speed_var': speed_variance,
'alcohol_score': alcohol_score,
'cognitive_score': cognitive_score,
}
}
if __name__ == "__main__":
np.random.seed(42)
normal_steering = np.cumsum(np.random.normal(0, 0.3, 1800))
normal_result = classify_impairment(
normal_steering, gaze_entropy=0.75,
pupil_diameter_var=0.08, speed_variance=5.0
)
print(f"正常驾驶: {normal_result['state']} (conf={normal_result['confidence']:.2f})")
drunk_steering = np.cumsum(np.random.normal(0, 0.3, 1800))
for i in range(0, 1800, 40):
drunk_steering[i:i+5] += np.random.normal(0, 3, 5)
drunk_result = classify_impairment(
drunk_steering, gaze_entropy=0.55,
pupil_diameter_var=0.10, speed_variance=9.0
)
print(f"酒驾: {drunk_result['state']} (conf={drunk_result['confidence']:.2f})")
combined_result = classify_impairment(
drunk_steering, gaze_entropy=0.50,
pupil_diameter_var=0.20, speed_variance=4.5
)
print(f"酒驾+分心: {combined_result['state']} (conf={combined_result['confidence']:.2f})")
|