V 1.0 - Only Prints

audio/audio_cross_correlation.py

@@ -0,0 +1,54 @@
+# for import needed
+import sys
+from pathlib import Path
+
+sys.path.append(str(Path(__file__).resolve().parent.parent))
+
+from utils import run
+### Normal Imports:
+import tempfile
+import subprocess
+import soundfile as sf
+import numpy as np
+from scipy.signal import correlate
+
+def extract_audio(file: Path, duration: float) -> tuple[np.ndarray, int]:
+    tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
+
+    subprocess.run([
+        "ffmpeg",
+        "-y",
+        "-i", str(file),
+        "-map", "0:a:0",        # first audio stream
+        "-ac", "1",             # mono
+        "-ar", "44100",         # fixed sample rate
+        "-t", str(duration),    # limit duration
+        tmp.name
+    ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+
+    data, sr = sf.read(tmp.name)
+    if not isinstance(data, np.ndarray):
+        raise ChildProcessError(f"Could not get signal from {file}")
+    if not isinstance(sr, int):
+        try:
+            sr = int(sr)
+        except:
+            ValueError(f"sr could not be read from {file}")
+    return data, sr
+
+def get_audio_offset(file1: Path, file2: Path, important_time: float) -> float:
+    sig1, sr1 = extract_audio(file1, important_time)
+    sig2, sr2 = extract_audio(file2, important_time)
+
+    if sr1 != sr2:
+        raise ValueError("Sample rates do not match")
+
+    # normalize signals
+    sig1 = sig1 - np.mean(sig1)
+    sig2 = sig2 - np.mean(sig2)
+
+    corr = correlate(sig1, sig2, mode="full")
+    lag = np.argmax(corr) - (len(sig2) - 1)
+
+    offset_seconds = float(lag / sr1)
+    return offset_seconds