detailed-audio-analysis/app.py at main · devbret/detailed-audio-analysis · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
import os
import json
import math
import librosa
import numpy as np
from tqdm import tqdm
import concurrent.futures

directory = "audio/"
output_file = "audio_analysis_enhanced.json"
HOP_LENGTH = 512

def finite_list(values, decimals):
    arr = np.round(np.asarray(values, dtype=float), decimals)
    out = arr.tolist()
    if not np.isfinite(arr).all():
        out = [v if math.isfinite(v) else None for v in out]
    return out

def series(values):
    return {"values": finite_list(values, 6)}

def events(times, values):
    return {
        "times": finite_list(times, 4),
        "values": finite_list(values, 6),
    }

def compute_tempo(y, sr, hop_length=HOP_LENGTH):
    onset_env = librosa.onset.onset_strength(y=y, sr=sr, hop_length=hop_length, aggregate=np.median)
    tempo, beat_frames = librosa.beat.beat_track(onset_envelope=onset_env, sr=sr, hop_length=hop_length, units='frames')
    beat_times = librosa.frames_to_time(beat_frames, sr=sr, hop_length=hop_length)
    instantaneous_tempo = 60.0 / np.diff(beat_times) if len(beat_times) > 1 else np.array([])
    return events(beat_times[:-1], instantaneous_tempo), float(np.atleast_1d(tempo)[0])

def analyze_audio(file_path, hop_length=HOP_LENGTH):
    y, sr = librosa.load(file_path)

    onset_env = librosa.onset.onset_strength(y=y, sr=sr, hop_length=hop_length, aggregate=np.median)
    novelty_vals = np.maximum(0, np.diff(onset_env))

    rms_energy = librosa.feature.rms(y=y, hop_length=hop_length)[0]

    onset_frames = librosa.onset.onset_detect(y=y, sr=sr, hop_length=hop_length)
    onset_times = librosa.frames_to_time(onset_frames, sr=sr, hop_length=hop_length)

    mfccs = librosa.feature.mfcc(y=y, sr=sr, hop_length=hop_length)
    chroma = librosa.feature.chroma_cqt(y=y, sr=sr, hop_length=hop_length)

    beats, overall_tempo = compute_tempo(y, sr, hop_length=hop_length)

    spectral_centroids = librosa.feature.spectral_centroid(y=y, sr=sr, hop_length=hop_length)[0]
    spectral_bandwidth = librosa.feature.spectral_bandwidth(y=y, sr=sr, hop_length=hop_length)[0]
    zero_crossings = librosa.feature.zero_crossing_rate(y=y, hop_length=hop_length)[0]
    spectral_contrast = librosa.feature.spectral_contrast(y=y, sr=sr, hop_length=hop_length)
    spectral_rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr, hop_length=hop_length)[0]

    mel_spectrogram = librosa.feature.melspectrogram(y=y, sr=sr, hop_length=hop_length)
    mel_db = librosa.power_to_db(mel_spectrogram, ref=np.max)

    tonnetz = librosa.feature.tonnetz(y=y, sr=sr, hop_length=hop_length)

    harmonic, percussive = librosa.effects.hpss(y)
    harm_energy = librosa.feature.rms(y=harmonic, hop_length=hop_length)[0]
    perc_energy = librosa.feature.rms(y=percussive, hop_length=hop_length)[0]
    harmonic_ratio = harm_energy / (harm_energy + perc_energy + 1e-8)

    S = np.abs(librosa.stft(y, hop_length=hop_length))
    flux_vals = np.sqrt(np.sum(np.diff(S, axis=1) ** 2, axis=0))

    f0, _, _ = librosa.pyin(
        y,
        fmin=librosa.note_to_hz('C2'),
        fmax=librosa.note_to_hz('C7'),
        sr=sr,
        frame_length=2048,
        hop_length=hop_length
    )

    centroid_delta = librosa.feature.delta(spectral_centroids)

    return {
        "sr": int(sr),
        "hop_length": hop_length,
        "duration": float(librosa.get_duration(y=y, sr=sr)),
        "overall_tempo": overall_tempo,
        "onsets": events(onset_times, np.ones(len(onset_times))),
        "onset_strength": series(onset_env),
        "novelty": series(novelty_vals),
        "loudness": series(rms_energy),
        "timbre": {f"mfcc{i+1}": series(m) for i, m in enumerate(mfccs)},
        "chroma": {f"chroma{i+1}": series(c) for i, c in enumerate(chroma)},
        "tempo": beats,
        "spectral_centroid": series(spectral_centroids),
        "spectral_bandwidth": series(spectral_bandwidth),
        "zero_crossing_rate": series(zero_crossings),
        "spectral_contrast": {f"contrast{i+1}": series(c) for i, c in enumerate(spectral_contrast)},
        "spectral_rolloff": series(spectral_rolloff),
        "mel_spectrogram": {f"mel{i+1}": series(m) for i, m in enumerate(mel_db)},
        "tonnetz": {f"tonnetz{i+1}": series(t) for i, t in enumerate(tonnetz)},
        "harmonics": series(harm_energy),
        "percussives": series(perc_energy),
        "harmonic_ratio": series(harmonic_ratio),
        "spectral_flux": series(flux_vals),
        "pitch": series(f0),
        "spectral_centroid_velocity": series(centroid_delta),
    }

def process_files(directory):
    analysis_results = {}
    files = [f for f in os.listdir(directory) if f.endswith((".mp3", ".wav"))]

    if not files:
        print(f"No audio files found in directory: {directory}")
        return analysis_results

    print(f"Found {len(files)} audio files in directory: {directory}")

    with concurrent.futures.ProcessPoolExecutor() as executor:
        futures = {executor.submit(analyze_audio, os.path.join(directory, filename)): filename for filename in files}
        for future in tqdm(concurrent.futures.as_completed(futures), total=len(futures), desc="Analyzing audio files"):
            filename = futures[future]
            try:
                analysis_results[filename] = future.result()
            except Exception as e:
                print(f"Error analyzing {filename}: {e}")

    return analysis_results

def numpy_to_python(obj):
    if isinstance(obj, np.ndarray):
        return obj.tolist()
    if isinstance(obj, np.generic):
        return float(obj)
    raise TypeError(f"Object of type {obj.__class__.__name__} is not JSON serializable")

def save_results(output_file, analysis_results):
    with open(output_file, 'w') as f:
        json.dump(analysis_results, f, separators=(",", ":"), default=numpy_to_python, allow_nan=False)
    print(f"Enhanced audio analysis completed. Results are saved in {output_file}")

if __name__ == "__main__":
    if not os.path.exists(directory):
        print(f"Directory {directory} does not exist.")
    else:
        analysis_results = process_files(directory)
        if analysis_results:
            save_results(output_file, analysis_results)