scratch/test_pitch_debug.py

import sys
import os
import numpy as np
import soundfile as sf


# Add CDAC_ASR to sys.path
sys.path.append("/home/mihir/Codes/CDAC_ASR")

from src.eval.ScoreCalcs import PronunciationScorer

def test_pitch():
    # Convert mp3 to wav
    mp3_path = "/home/mihir/Codes/CDAC_ASR/product/backend/cache/tts/cf673f7ee88828c9fb8f6acf2cb08403_normal.mp3"
    wav_path = "/home/mihir/Codes/CDAC_ASR/scratch/test_debug.wav"
    
    import subprocess
    cmd = ["ffmpeg", "-y", "-i", mp3_path, "-ar", "16000", "-ac", "1", wav_path]
    subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
    
    speech, sr = sf.read(wav_path)
    print(f"Loaded audio: shape={speech.shape}, sr={sr}")
    
    scorer = PronunciationScorer()
    
    # Dummy alignment and phoneme times
    # Let's say we have 5 phonemes
    n_phonemes = 5
    duration = len(speech) / sr
    step = duration / n_phonemes
    times = [(i*step, (i+1)*step) for i in range(n_phonemes)]
    
    aligned_pairs = [("a", "a")] * n_phonemes
    
    # Extract contours
    pred_contours = scorer._extract_pitch_contour(speech, sr, times)
    print(f"Pred contours: {len(pred_contours)}")
    for i, c in enumerate(pred_contours):
        print(f"  Phoneme {i}: len={len(c)}, mean={c.mean() if len(c) > 0 else 'empty'}")
        
    res = scorer.pitch_score(speech, speech, sr, aligned_pairs, times, times)
    print("Pitch score result:", res)

if __name__ == "__main__":
    test_pitch()