从原理到实践详解Python音频录制与分析系统的实现

CHUNK = 1024          # 每次读取的音频块大小（帧数）
FORMAT = pyaudio.paInt16  # 采样格式（16位有符号整数）
CHANNELS = 1          # 单声道录制
RATE = 16000          # 采样率16kHz（适合语音识别）

sequenceDiagram
    participant User
    participant PyAudio
    participant WaveFile
    
    User->>PyAudio: 初始化音频流
    loop 每次读取CHUNK
        PyAudio->>PyAudio: 从麦克风读取数据
        PyAudio->>WaveFile: 存储音频帧
    end
    User->>PyAudio: 停止流
    PyAudio->>WaveFile: 写入文件头信息

p = pyaudio.PyAudio()  # 创建PyAudio实例
 
stream = p.open(
    format=FORMAT,
    channels=CHANNELS,
    rate=RATE,
    input=True,          # 输入模式（录音）
    frames_per_buffer=CHUNK
)  # 打开音频流
 
frames = []
for _ in range(0, int(RATE / CHUNK * duration)):
    data = stream.read(CHUNK)  # 读取音频数据
    frames.append(data)        # 存储到列表

with wave.open(filename, 'wb') as wf:
    wf.setnchannels(CHANNELS)  # 设置声道数
    wf.setsampwidth(p.get_sample_size(FORMAT))  # 采样宽度
    wf.setframerate(RATE)      # 采样率
    wf.writeframes(b''.join(frames))  # 写入所有帧

def extract_wav_metadata(filepath):
    with wave.open(filepath, 'rb') as wf:
        return {
            "n_channels": wf.getnchannels(),  # 声道数
            "sample_width": wf.getsampwidth(), # 采样宽度（字节）
            "framerate": wf.getframerate(),    # 采样率
            "n_frames": wf.getnframes(),       # 总帧数
            "duration": wf.getnframes() / wf.getframerate()  # 时长
        }

{
  "n_channels": 1,
  "sample_width": 2,
  "framerate": 16000,
  "n_frames": 80000,
  "duration": 5.0
}

def compute_average_volume(frames, sample_width):
    rms_values = [audioop.rms(frame, sample_width) for frame in frames]
    return sum(rms_values) / len(rms_values) if rms_values else 0

def rms_to_decibel(rms):
    return 20 * math.log10(rms) if rms > 0 else -float('inf')

def simulate_noise_analysis(frames, sample_width):
    avg_rms = compute_average_volume(frames, sample_width)
    db = rms_to_decibel(avg_rms)
    level = "安静" if db < 30 else "适中" if db < 50 else "嘈杂"
    return {"rms": avg_rms, "db": db, "level": level}

def simulate_noise_analysis(frames, sample_width):
    avg_rms = compute_average_volume(frames, sample_width)
    db = rms_to_decibel(avg_rms)
    level = "安静" if db < 30 else "适中" if db < 50 else "嘈杂"
    return {"rms": avg_rms, "db": db, "level": level}

def placeholder_audio_quality_score(meta):
    base = 100
    if meta["n_channels"] != 1: base -= 20  # 非单声道扣分
    if meta["sample_width"] < 2: base -= 10 # 16位以下扣分
    return base

def generate_timestamped_name(base="record"):
    return f"{base}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.wav"

def analyze_directory(directory=".", suffix=".wav"):
    return [(f, extract_wav_metadata(os.path.join(directory, f))) 
            for f in os.listdir(directory) 
            if f.endswith(suffix)]

def simulate_wave_energy_curve(duration, rate=16000):
    time = np.linspace(0, duration, int(rate * duration))
    energy = np.abs(np.sin(2 * np.pi * time))  # 模拟正弦波能量
    return list(zip(time.tolist(), energy.tolist()))

def fake_noise_profile():
    return {
        "静音占比": "12%",      # 静音段比例
        "人声强度": "中",       # 人声能量水平
        "高频干扰": "无",       # 高频噪声
        "能量峰值位置": "2.1s"  # 最大能量位置
    }

try:
    with wave.open(filepath, 'rb') as wf:
        wf.getparams()
except wave.Error as e:
    logging.error(f"WAV文件解析失败: {e}")
    return False

# 根据不同硬件调整CHUNK大小
CHUNK = 512   # 低延迟
CHUNK = 2048  # 高吞吐

# 分批处理大文件
while True:
    data = stream.read(CHUNK)
    if not data: break
    process_frame(data)

def callback(in_data, frame_count, time_info, status):
    analyze_frame(in_data)  # 实时分析
    return (in_data, pyaudio.paContinue)

def preprocess_for_asr(filepath):
    meta = extract_wav_metadata(filepath)
    if meta["framerate"] != 16000:
        resample_audio(filepath, 16000)  # 重采样到16kHz
    if meta["n_channels"] > 1:
        convert_to_mono(filepath)        # 转单声道
    normalize_volume(filepath)           # 音量标准化

class AudioQualityTester:
    def __init__(self):
        self.thresholds = {
            'noise_db': 40,      # 最大允许噪声
            'duration_min': 1.0,  # 最短时长
            'clipping': False     # 是否允许削波
        }
    
    def test_file(self, filepath):
        report = {}
        meta = extract_wav_metadata(filepath)
        report.update(self.check_duration(meta))
        report.update(self.check_noise_level(filepath))
        report['passed'] = all(report.values())
        return report