import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA

class EDrunkDrivingSystem:
    """
    e-Nose辅助多信号融合酒驾检测系统
    
    论文: Sensors and Actuators B: Chemical (2025)
    灵敏度: 97.14%, 准确率: 98.67%
    """
    
    # 8个MOS气体传感器配置
    MOS_SENSORS = {
        "MQ-3":  {"target": "酒精(C2H5OH)", "range": "0.05-10mg/L", "sensitivity": 0.5},
        "MQ-135": {"target": "氨气/CO2", "range": "10-1000ppm", "sensitivity": 0.3},
        "MQ-2":  {"target": "烟雾/可燃气体", "range": "100-10000ppm", "sensitivity": 0.4},
        "MQ-7":  {"target": "一氧化碳CO", "range": "20-2000ppm", "sensitivity": 0.2},
        "MQ-4":  {"target": "甲烷CH4", "range": "200-10000ppm", "sensitivity": 0.3},
        "MQ-8":  {"target": "氢气H2", "range": "100-1000ppm", "sensitivity": 0.4},
        "MQ-9":  {"target": "CO/可燃气体", "range": "100-10000ppm", "sensitivity": 0.3},
        "MQ-131": {"target": "臭氧O3", "range": "10-1000ppb", "sensitivity": 0.2}
    }
    
    def __init__(self):
        self.scaler = StandardScaler()
        self.pca = PCA(n_components=12)
        self.classifier = RandomForestClassifier(
            n_estimators=200, max_depth=15, random_state=42
        )
    
    def extract_gas_features(self, sensor_readings):
        """
        从8个MOS传感器读数中提取特征
        
        Args:
            sensor_readings: shape=(N, 8), 各传感器响应值
        
        Returns:
            features: 提取的特征向量
        """
        features = []
        for i in range(sensor_readings.shape[1]):
            sig = sensor_readings[:, i]
            # 时域特征
            features.extend([
                np.mean(sig), np.std(sig), np.max(sig),
                np.min(sig), np.median(sig),
                np.percentile(sig, 75) - np.percentile(sig, 25),  # IQR
                # 频域特征
                np.mean(np.abs(np.diff(sig))),  # 变化率
                np.max(np.abs(np.diff(sig))),   # 最大变化率
                # 稳态特征
                sig[-10:].mean(),  # 末段均值
                sig[-10:].std(),   # 末段稳定性
            ])
        
        return np.array(features)  # 80维特征 (8×10)
    
    def extract_physio_features(self, physio_data):
        """
        生理信号特征提取
        心率、皮肤电反应(GSR)、体温
        """
        features = []
        for signal in physio_data:
            features.extend([
                np.mean(signal), np.std(signal),
                np.max(signal), np.min(signal)
            ])
        return np.array(features)
    
    def fuse_signals(self, gas_features, physio_features, visual_features):
        """多信号特征级融合"""
        # PCA降维各模态
        gas_reduced = self.pca.fit_transform(
            gas_features.reshape(1, -1)
        ).flatten()
        
        # 拼接所有模态
        fused = np.concatenate([
            gas_reduced,       # 12维
            physio_features,    # 12维 (3信号×4)
            visual_features    # 8维
        ])
        
        return fused  # 32维融合特征
    
    def predict(self, fused_features):
        """
        判断是否酒驾
        
        Returns:
            prediction: 0=正常, 1=醉酒
            confidence: 置信度
        """
        proba = self.classifier.predict_proba(
            fused_features.reshape(1, -1)
        )[0]
        
        return int(proba[1] > 0.5), proba[1]


# 测试代码
if __name__ == "__main__":
    system = EDrunkDrivingSystem()
    
    # 模拟8个MOS传感器30秒读数（采样率1Hz）
    np.random.seed(42)
    normal_gas = np.random.normal(0.3, 0.05, (30, 8))
    drunk_gas = np.random.normal(0.7, 0.1, (30, 8))  # 酒精浓度升高
    
    gas_features_normal = system.extract_gas_features(normal_gas)
    gas_features_drunk = system.extract_gas_features(drunk_gas)
    
    print(f"正常驾驶气体特征维度: {gas_features_normal.shape}")
    print(f"醉酒驾驶气体特征维度: {gas_features_drunk.shape}")
    print(f"正常驾驶MQ-3均值: {np.mean(normal_gas[:, 0]):.3f}")
    print(f"醉酒驾驶MQ-3均值: {np.mean(drunk_gas[:, 0]):.3f}")

class MultiSignalFusion:
    """三种融合策略对比"""
    
    @staticmethod
    def early_fusion(gas, physio, visual):
        """早期融合（特征级）- 论文最优方案"""
        return np.concatenate([gas, physio, visual])
    
    @staticmethod
    def late_fusion(pred_gas, pred_physio, pred_visual):
        """晚期融合（决策级）"""
        # 加权平均
        weights = [0.5, 0.3, 0.2]  # 气体权重最高
        return np.average([pred_gas, pred_physio, pred_visual], 
                         weights=weights)
    
    @staticmethod
    def hybrid_fusion(gas, physio, visual):
        """混合融合"""
        # 先气体+生理特征级融合
        early = np.concatenate([gas, physio])
        # 再与视觉决策级融合
        return early  # 简化示意

# 融合策略性能对比
fusion_results = {
    "Early fusion (特征级)": {
        "accuracy": 0.9867,
        "sensitivity": 0.9714,
        "specificity": 0.9921,
        "f1_score": 0.9812
    },
    "Late fusion (决策级)": {
        "accuracy": 0.9456,
        "sensitivity": 0.9286,
        "specificity": 0.9521,
        "f1_score": 0.9375
    },
    "Hybrid fusion": {
        "accuracy": 0.9678,
        "sensitivity": 0.9571,
        "specificity": 0.9721,
        "f1_score": 0.9623
    }
}

class HybridAlcoholDetection:
    """
    视觉+e-Nose混合酒驾检测架构
    
    Smart Eye视觉（行为分析）+ e-Nose（气体检测）
    = 多维度互补验证
    """
    
    def __init__(self):
        self.visual_dms = SmartEyeAlcoholModule()  # 视觉模块
        self.enose_module = ENoseModule()           # 气体模块
        self.physio_module = PhysioModule()         # 生理模块
        self.fusion_weights = {"visual": 0.35, "gas": 0.45, "physio": 0.20}
    
    def assess_impairment(self, sensor_data):
        """
        多模态酒精损伤评估
        
        Returns:
            impairment_level: 0-100
            action: 干预策略
        """
        visual_score = self.visual_dms.analyze(sensor_data["video"])
        gas_score = self.enose_module.detect(sensor_data["gas"])
        physio_score = self.physio_module.analyze(sensor_data["physio"])
        
        # 加权融合
        total = (
            visual_score * self.fusion_weights["visual"] +
            gas_score * self.fusion_weights["gas"] +
            physio_score * self.fusion_weights["physio"]
        )
        
        if total > 80:
            return total, "ENGINE_LOCK"
        elif total > 60:
            return total, "WARNING_LEVEL_2"
        elif total > 40:
            return total, "WARNING_LEVEL_1"
        else:
            return total, "NORMAL"