"""音频解码模块:从视频/音频文件中提取单声道时域信号。 依赖系统已安装 ffmpeg(需在 PATH 中)。 """ import subprocess import tempfile import os from math import gcd from pathlib import Path import numpy as np import soundfile as sf DEFAULT_SR = 44100 _NATIVE_AUDIO_EXTS = {".wav", ".flac", ".ogg", ".aiff", ".aif"} def decode_audio(file_path: str, target_sr: int = DEFAULT_SR) -> tuple: """ 从任意音频/视频文件解码,返回单声道 float32 时域信号。 Parameters ---------- file_path : str 输入文件路径,支持 .mp4 .mov .avi .mkv .mp3 .m4a .wav .flac 等。 target_sr : int 目标采样率,默认 44100 Hz。 Returns ------- samples : np.ndarray shape (N,) float32,范围约 [-1, 1] sample_rate : int 实际采样率(== target_sr) """ file_path = str(file_path) ext = Path(file_path).suffix.lower() if ext not in _NATIVE_AUDIO_EXTS: wav_path = _ffmpeg_to_wav(file_path, target_sr) tmp = True else: wav_path = file_path tmp = False try: samples, sr = sf.read(wav_path, dtype="float32", always_2d=False) finally: if tmp and os.path.exists(wav_path): os.remove(wav_path) # 多声道 → 单声道 if samples.ndim > 1: samples = samples.mean(axis=1) # 重采样 if sr != target_sr: samples = _resample(samples, sr, target_sr) return samples.astype(np.float32), target_sr def _ffmpeg_to_wav(src: str, sr: int) -> str: """用 ffmpeg 将任意媒体转为单声道 WAV,返回临时文件路径。""" tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False) tmp_path = tmp.name tmp.close() cmd = [ "ffmpeg", "-y", "-i", src, "-ac", "1", "-ar", str(sr), "-vn", tmp_path, ] result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE) if result.returncode != 0: raise RuntimeError( f"ffmpeg 解码失败:\n{result.stderr.decode(errors='replace')}" ) return tmp_path def _resample(samples: np.ndarray, orig_sr: int, target_sr: int) -> np.ndarray: """整数比率重采样(scipy polyphase)。""" from scipy.signal import resample_poly g = gcd(orig_sr, target_sr) return resample_poly(samples, target_sr // g, orig_sr // g).astype(np.float32)