"""
TI AWRL6432 60GHz CPD 方案

关键特性:
- 单芯片集成(4发4收)
- 边缘AI加速(可 onboard处理)
- 功耗 <200mW(适合熄火后长时间监测)
"""

awrl6432_spec = {
    "芯片": "AWRL6432",
    "频率": "60-64GHz",
    "带宽": "4GHz",
    "天线配置": "4发4收(16虚拟通道)",
    "距离分辨率": "3.75cm",
    "速度分辨率": "0.3km/h",
    "最大距离": "6m(车内足够)",
    "功耗": "200mW(活跃), <10mW(待机)",
    "ADC": "12-bit, 8MHz IF",
    "车规": "AEC-Q100 Grade 2",
    "SDK": "mmWave SDK + Edge AI SDK",
    "评估板": "AWRL6432BOOST",
    "价格": "~$8(量产预估)",
    "CPD验证": "TI已发布demo视频(2026.07)",
    "优势": [
        "成本最低的mmWave方案",
        "TI Edge AI SDK支持onboard推理",
        "功耗低，适合熄火后电池供电",
        "TI Model Zoo有CPD参考模型"
    ],
    "劣势": [
        "60GHz穿透座椅能力不如UWB",
        "距离分辨率不如79GHz",
        "实际车内多径环境需充分验证"
    ]
}

print("=== TI AWRL6432 CPD方案 ===")
for k, v in awrl6432_spec.items():
    if isinstance(v, list):
        print(f"{k}:")
        for item in v:
            print(f"  - {item}")
    else:
        print(f"{k}: {v}")

"""
Infineon BGT60ATR24AIP 60GHz 车载雷达

关键特性:
- AIP(Antenna-in-Package)设计,紧凑封装
- Mouser已开始分销(2026.08)
- 支持生命体征监测
"""

bgt60_spec = {
    "芯片": "BGT60ATR24AIP",
    "频率": "60GHz",
    "带宽": "4GHz",
    "天线配置": "2发4收, AIP封装",
    "距离分辨率": "3.75cm",
    "最大距离": "6m",
    "功耗": "250mW",
    "封装": "AIP(天线集成在封装内)",
    "车规": "AEC-Q100",
    "SDK": "Infineon Radar SDK + BGT60 SDK",
    "分销": "Mouser (2026.08起)",
    "CPD验证": "已有参考设计",
    "生命体征": "支持呼吸频率检测(0.1-0.5mm位移)",
    "优势": [
        "AIP设计无需额外天线设计",
        "已量产分销,供应链成熟",
        "Infineon有完整参考设计"
    ],
    "劣势": [
        "功耗略高于TI方案",
        "无onboard AI加速",
        "需外接MCU做ML推理"
    ]
}

"""
NXP UWB SR150 CPD方案

核心优势: 复用UWB数字钥匙基础设施
- 车辆已安装UWB锚点用于数字钥匙
- 同一芯片可做CPD,零额外硬件成本
"""

sr150_spec = {
    "芯片": "NXP SR150/NCJ29D5B",
    "频率": "6.5-8GHz(UWB)",
    "带宽": "500MHz",
    "天线配置": "取决于车辆UWB锚点布局",
    "距离分辨率": "~30cm(不如mmWave)",
    "最大距离": "3m(车内足够)",
    "功耗": "<50mW",
    "车规": "AEC-Q100 Grade 3",
    "SDK": "NXP UWB SDK + Radar Processing",
    "核心优势": "复用UWB数字钥匙基础设施,零额外硬件",
    "生命体征": "支持呼吸检测(相位变化法)",
    "CPD验证": "NXP已发布参考设计(2026)",
    "优势": [
        "零额外硬件成本(复用UWB锚点)",
        "功耗最低(<50mW)",
        "穿透座椅能力强(低频)",
        "可同时做数字钥匙+CPD+手势识别"
    ],
    "劣势": [
        "距离分辨率差(30cm vs 3.75cm)",
        "无法精确定位儿童位置",
        "UWB锚点位置为钥匙优化,非CPD优化",
        "需验证后排/后备箱覆盖"
    ]
}

"""
雷达呼吸检测核心算法

原理:
- 人体呼吸时胸壁有0.1-0.5mm的微小位移
- 雷达可检测到这种微多普勒相位变化
- 通过相位提取+滤波获得呼吸频率

适用于: 60GHz/79GHz mmWave + UWB
"""

import numpy as np
from scipy.signal import butter, filtfilt, find_peaks

def radar_breathing_detection(
    phase_data: np.ndarray,
    sample_rate: float = 20,  # Hz
    duration_sec: float = 30  # 30秒窗口
) -> dict:
    """
    雷达相位数据提取呼吸频率
    
    Args:
        phase_data: 雷达ADC相位数据, shape=(N,)
        sample_rate: 帧率(Hz)
        duration_sec: 分析窗口长度
    
    Returns:
        result: {
            'breathing_rate': 呼吸频率(次/分钟),
            'confidence': 置信度,
            'is_present': 是否检测到活体
        }
    """
    n_samples = int(sample_rate * duration_sec)
    if len(phase_data) < n_samples:
        n_samples = len(phase_data)
    
    signal = phase_data[:n_samples]
    
    # 1. 去趋势(去除直流分量和慢漂移)
    signal = signal - np.mean(signal)
    
    # 2. 带通滤波: 呼吸频率 0.1-0.5 Hz (6-30次/分钟)
    nyq = sample_rate / 2
    low = 0.1 / nyq
    high = 0.5 / nyq
    b, a = butter(4, [low, high], btype='band')
    filtered = filtfilt(b, a, signal)
    
    # 3. FFT分析呼吸频率
    freqs = np.fft.rfftfreq(len(filtered), 1/sample_rate)
    fft_mag = np.abs(np.fft.rfft(filtered))
    
    # 只看呼吸频率范围
    breath_mask = (freqs >= 0.1) & (freqs <= 0.5)
    breath_freqs = freqs[breath_mask]
    breath_mag = fft_mag[breath_mask]
    
    if len(breath_mag) == 0:
        return {'breathing_rate': 0, 'confidence': 0, 'is_present': False}
    
    # 主频 = 呼吸频率
    peak_idx = np.argmax(breath_mag)
    breathing_freq = breath_freqs[peak_idx]
    breathing_rate = breathing_freq * 60  # 转 次/分钟
    
    # 置信度 = 峰值/总能量比
    total_energy = np.sum(breath_mag)
    peak_energy = breath_mag[peak_idx]
    confidence = peak_energy / (total_energy + 1e-8)
    
    # 判断是否检测到活体
    # 正常呼吸: 12-20次/分钟(0.2-0.33Hz)
    # 婴儿呼吸: 30-60次/分钟(0.5-1.0Hz)
    is_present = confidence > 0.3 and breathing_rate > 5
    
    return {
        'breathing_rate': round(breathing_rate, 1),
        'confidence': round(confidence, 3),
        'is_present': is_present,
        'breath_spectrum': list(zip(breath_freqs.tolist(), breath_mag.tolist()))
    }


# ========== 测试 ==========
if __name__ == "__main__":
    np.random.seed(42)
    
    # 模拟有儿童(呼吸频率25次/分钟)
    t = np.arange(0, 30, 0.05)  # 30秒, 20Hz
    child_breathing = 0.5 * np.sin(2 * np.pi * 25/60 * t)  # 25次/分钟
    noise = np.random.normal(0, 0.05, len(t))
    phase_with_child = child_breathing + noise
    
    result = radar_breathing_detection(phase_with_child, sample_rate=20, duration_sec=30)
    print(f"=== 有儿童场景 ===")
    print(f"呼吸频率: {result['breathing_rate']} 次/分钟")
    print(f"置信度: {result['confidence']}")
    print(f"检测到活体: {result['is_present']}")
    
    # 模拟无儿童(纯噪声)
    t = np.arange(0, 30, 0.05)
    noise_only = np.random.normal(0, 0.02, len(t))
    
    result2 = radar_breathing_detection(noise_only, sample_rate=20, duration_sec=30)
    print(f"\n=== 无儿童场景 ===")
    print(f"呼吸频率: {result2['breathing_rate']} 次/分钟")
    print(f"置信度: {result2['confidence']}")
    print(f"检测到活体: {result2['is_present']}")