import numpy as np
import matplotlib.pyplot as plt

# 模拟LFN振幅数据（基于论文描述的趋势）
conditions = ['Sign', 'Brake', 'signBrake']
young_lfn = [-2.1, -3.5, -5.2]   # μV, 年轻人
old_lfn = [-1.8, -2.3, -2.9]     # μV, 老年人

x = np.arange(len(conditions))
width = 0.35

fig, ax = plt.subplots(figsize=(10, 6))
bars1 = ax.bar(x - width/2, young_lfn, width, label='Younger (19-25)', color='steelblue')
bars2 = ax.bar(x + width/2, old_lfn, width, label='Older (65-74)', color='coral')

ax.set_ylabel('LFN Amplitude (μV)')
ax.set_title('LFN Condition Differentiation by Age Group')
ax.set_xticks(x)
ax.set_xticklabels(conditions)
ax.legend()
ax.axhline(y=0, color='black', linewidth=0.5)

# 关键发现标注
ax.annotate('老年人LFN分化弱化\n→预备性控制下降', 
            xy=(2.35, -2.9), xytext=(2.5, -1),
            arrowprops=dict(arrowstyle='->', color='red'),
            fontsize=10, color='red')

plt.tight_layout()
plt.savefig('lfn_age_comparison.png', dpi=150)
plt.close()
print("LFN对比图已生成")

# 关键指标计算
young_diff = young_lfn[2] - young_lfn[0]  # signBrake - Sign
old_diff = old_lfn[2] - old_lfn[0]
print(f"年轻人 LFN 条件间分化: {young_diff:.1f} μV")
print(f"老年人 LFN 条件间分化: {old_diff:.1f} μV")
print(f"分化削弱比例: {(1 - old_diff/young_diff)*100:.0f}%")

import numpy as np

# 模拟时频分析数据（基于论文描述）
bands = {
    'Frontal Theta (4-8Hz)': {'young': 1.0, 'old': 0.72, 'function': '认知控制/冲突监测'},
    'Parietal Alpha (8-13Hz)': {'young': 1.0, 'old': 0.68, 'function': '注意力门控/抑制'},
    'Central Beta (13-30Hz)': {'young': 1.0, 'old': 0.61, 'function': '运动准备/执行'},
}

print("=" * 70)
print("频段功率年龄差异（归一化到年轻人=1.0）")
print("=" * 70)
print(f"{'频段':<25} {'年轻人':>8} {'老年人':>8} {'降幅':>8} {'功能'}")
print("-" * 70)
for band, data in bands.items():
    reduction = (1 - data['old']/data['young']) * 100
    print(f"{band:<25} {data['young']:>8.2f} {data['old']:>8.2f} {reduction:>7.1f}% {data['function']}")
    
print("\n关键发现：")
print("1. 中央区β功率在Age×Condition交互中效应最强")
print("2. 老年人在所有条件下均表现为频段功率降低")
print("3. 年轻人的频段功率无显著条件差异（更灵活的调节）")
print("4. 老年人的β功率条件间差异显著→运动准备策略不同")

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.model_selection import cross_val_score
import numpy as np

# 模拟MVPA年龄解码（基于论文描述的方法）
def simulate_mvpa_decoding():
    """
    MVPA: 使用时频功率特征解码年龄组别
    方法: LDA + 5-fold cross-validation
    时间窗口: -200ms 到 800ms (相对刺激)
    """
    # 模拟时间点和准确率
    time_points = np.linspace(-200, 800, 50)  # ms
    # 刺激前: 接近随机水平 (50%)
    # 刺激后: 显著高于随机
    accuracy = np.where(
        time_points < 0,
        0.50 + np.random.normal(0, 0.02, len(time_points)),
        0.50 + 0.15 * (1 - np.exp(-time_points / 100))  # 上升曲线
    )
    
    # 关键时间点
    onset_time = 0  # 刺激开始
    above_chance_time = 50  # 50ms后开始显著
    peak_accuracy = max(accuracy)
    
    print(f"MVPA 年龄解码结果：")
    print(f"  刺激前准确率: ~50% (随机水平)")
    print(f"  刺激后准确率: 显著提升")
    print(f"  峰值准确率: {peak_accuracy:.1%}")
    print(f"  显著高于随机的时间: ~{above_chance_time}ms 后")
    
    # 跨条件解码
    conditions = ['Sign', 'Brake', 'signBrake']
    print(f"\n跨条件解码（在一条件训练，另一条件测试）：")
    for train_c in conditions:
        for test_c in conditions:
            if train_c != test_c:
                # 模拟跨条件准确率
                acc = 0.58 + np.random.normal(0, 0.02)
                print(f"  {train_c}→{test_c}: {acc:.1%} {'✅' if acc > 0.55 else '❌'}")
    
    return time_points, accuracy

time_points, accuracy = simulate_mvpa_decoding()
print("\n结论: 年龄相关神经特征跨条件泛化→分布式年龄标记")

class AgeAdaptiveDMS:
    """年龄自适应 DMS 配置"""
    
    # 年龄相关参数调整
    AGE_THRESHOLDS = {
        'young': {  # 18-30
            'perclos_threshold': 0.15,      # PERCLOS 闭眼比例阈值
            'gaze_off_threshold': 3.0,       # 视线偏离秒数
            'blink_rate_min': 10,            # 正常眨眼频率下限
            'blink_rate_max': 20,            # 正常眨眼频率上限
            'brake_reaction_time': 0.5,      # 正常刹车反应时间
        },
        'middle': {  # 30-60
            'perclos_threshold': 0.15,
            'gaze_off_threshold': 3.0,
            'blink_rate_min': 12,
            'blink_rate_max': 22,
            'brake_reaction_time': 0.6,
        },
        'older': {  # 65+
            'perclos_threshold': 0.12,       # 更低阈值（眨眼模式不同）
            'gaze_off_threshold': 2.5,       # 更短容忍（反应更慢）
            'blink_rate_min': 15,            # 眨眼更频繁但更短
            'blink_rate_max': 25,
            'brake_reaction_time': 0.8,      # 反应时间更长
        }
    }
    
    def get_config(self, estimated_age):
        """根据估计年龄获取DMS配置"""
        if estimated_age < 30:
            return self.AGE_THRESHOLDS['young']
        elif estimated_age < 65:
            return self.AGE_THRESHOLDS['middle']
        else:
            return self.AGE_THRESHOLDS['older']
    
    def assess_cognitive_load(self, driver_state):
        """
        认知负荷评估（基于论文发现的ERP模式）
        
        老年驾驶员模式：
        - 早期注意力补偿增强（N1↑）
        - 后期评估减弱（P3↓）
        - 预备控制灵活性下降（LFN分化弱化）
        
        DMS可观测的代理指标：
        - 眨眼时长变化
        - 眨眼振幅-速度比
        - 方向盘转角稳定性
        """
        config = self.get_config(driver_state.estimated_age)
        
        # 眨眼模式分析
        blink_warning = 0
        if driver_state.avg_blink_duration < config['blink_duration_min']:
            blink_warning += 0.3  # 短眨眼可能暗示注意力问题
        
        # PERCLOS检查
        if driver_state.perclos > config['perclos_threshold']:
            blink_warning += 0.5
        
        # 方向盘稳定性（老年人在双任务下方向盘更不稳定）
        steering_instability = abs(driver_state.steering_angle_std)
        if steering_instability > config['steering_threshold']:
            blink_warning += 0.2
        
        return {
            'cognitive_load': min(blink_warning, 1.0),
            'recommendation': 'reduce_distraction' if blink_warning > 0.5 else 'normal'
        }