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
| import numpy as np
from scipy.signal import welch
class RatioBasedThresholding:
"""
比值阈值标定（RBT）
论文核心方法：
1. 计算清醒基线θ/β比值
2. 动态计算当前θ/β比值
3. 基于基线偏差生成疲劳标签
"""
def __init__(self, fs: int = 200,
theta_band=(4, 8),
beta_band=(13, 30)):
self.fs = fs
self.theta_band = theta_band
self.beta_band = beta_band
def compute_theta_beta_ratio(self, eeg: np.ndarray) -> float:
"""
计算θ/β功率比值
Args:
eeg: [n_channels, n_samples] EEG信号
Returns:
ratio: θ/β功率比值
"""
eeg = eeg - eeg.mean(axis=0)
freqs, psd = welch(eeg, fs=self.fs, nperseg=512)
theta_mask = (freqs >= self.theta_band[0]) & \
(freqs < self.theta_band[1])
beta_mask = (freqs >= self.beta_band[0]) & \
(freqs < self.beta_band[1])
theta_power = np.sum(psd[:, theta_mask], axis=1).mean()
beta_power = np.sum(psd[:, beta_mask], axis=1).mean()
return theta_power / (beta_power + 1e-8)
def calibrate_baseline(self, alert_eeg: np.ndarray,
window_sec: int = 10) -> float:
"""
从清醒状态EEG标定个体基线
Args:
alert_eeg: 清醒状态EEG数据
window_sec: 窗口长度
Returns:
baseline_ratio: 基线θ/β比值
"""
window_samples = window_sec * self.fs
n_windows = len(alert_eeg[0]) // window_samples
ratios = []
for i in range(n_windows):
window = alert_eeg[:, i*window_samples:(i+1)*window_samples]
ratios.append(self.compute_theta_beta_ratio(window))
baseline = np.mean(ratios)
baseline_std = np.std(ratios)
return baseline, baseline_std
def generate_labels(self, eeg: np.ndarray,
baseline: float,
baseline_std: float,
threshold_std: float = 1.5) -> np.ndarray:
"""
基于RBT生成疲劳标签
Args:
eeg: [n_channels, n_samples]
baseline: 基线θ/β比值
baseline_std: 基线标准差
threshold_std: 阈值标准差倍数
Returns:
labels: [n_windows] 0=清醒, 1=疲劳
"""
window_samples = 10 * self.fs
n_windows = len(eeg[0]) // window_samples
labels = np.zeros(n_windows)
ratios = []
for i in range(n_windows):
window = eeg[:, i*window_samples:(i+1)*window_samples]
ratio = self.compute_theta_beta_ratio(window)
ratios.append(ratio)
if ratio > baseline + threshold_std * baseline_std:
labels[i] = 1
return labels, ratios
if __name__ == "__main__":
rbt = RatioBasedThresholding(fs=200)
np.random.seed(42)
alert_eeg = np.random.randn(4, 200 * 60)
alert_eeg += 2.0 * np.sin(2 * np.pi * 20 *
np.arange(200*60) / 200)
baseline, baseline_std = rbt.calibrate_baseline(alert_eeg)
print(f"基线θ/β比值: {baseline:.4f} ± {baseline_std:.4f}")
fatigue_eeg = np.random.randn(4, 200 * 60)
fatigue_eeg += 3.0 * np.sin(2 * np.pi * 6 *
np.arange(200*60) / 200)
labels, ratios = rbt.generate_labels(
fatigue_eeg, baseline, baseline_std
)
print(f"疲劳窗口比例: {labels.mean():.1%}")
print(f"平均θ/β比值: {np.mean(ratios):.4f}")
print(f"基线+1.5σ阈值: {baseline + 1.5 * baseline_std:.4f}")
|