File size: 4,833 Bytes
b72f14c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27b0097
b72f14c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27b0097
b72f14c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import numpy as np
import librosa
import scipy.signal as sps


def compute_spectral_analysis(y, sr, n_fft=4096):
    """Comprehensive spectral analysis tuned for speech QC."""

    hop_length = n_fft // 4

    # ============================================================
    # STFT → Magnitude + dB Conversion
    # ============================================================
    S = np.abs(librosa.stft(
        y,
        n_fft=n_fft,
        hop_length=hop_length,
        window="hann"
    ))

    freqs = np.linspace(0, sr / 2, S.shape[0])

    # Convert amplitude to dB scale
    S_db = librosa.amplitude_to_db(S, ref=np.max)

    # ============================================================
    # 90th Percentile Energy Envelope
    # ============================================================
    S_power = S ** 2
    energy = np.percentile(S_power, 90, axis=1) + 1e-20
    total_energy = float(np.sum(energy))

    cum_energy = np.cumsum(energy)

    roll85_idx = np.searchsorted(cum_energy, 0.85 * total_energy)
    roll95_idx = np.searchsorted(cum_energy, 0.95 * total_energy)

    freq_at_85 = float(freqs[min(roll85_idx, len(freqs) - 1)])
    freq_at_95 = float(freqs[min(roll95_idx, len(freqs) - 1)])

    # ============================================================
    # Updated HF Envelope: 90th percentile of dB
    # ============================================================
    mean_db_per_bin = np.percentile(S_db, 90, axis=1)

    peak_db = float(np.max(S_db))
    threshold_db = peak_db - 60

    non_silent_bins = np.where(mean_db_per_bin > threshold_db)[0]
    highest_freq = float(freqs[non_silent_bins[-1]]) if non_silent_bins.size else 0.0

    # ============================================================
    # Speech-Centric Band Energy Distribution
    # ============================================================
    def band_energy(low, high):
        i1 = np.searchsorted(freqs, low)
        i2 = np.searchsorted(freqs, high)
        return float(100 * np.sum(energy[i1:i2]) / total_energy)

    def band_energy_above(f):
        idx = np.searchsorted(freqs, f)
        return float(100 * np.sum(energy[idx:]) / total_energy)

    energy_stats = {
        "below_100hz": band_energy(0, 100),
        "100_500hz": band_energy(100, 500),
        "500_2khz": band_energy(500, 2000),
        "2k_8khz": band_energy(2000, 8000),
        "8k_12khz": band_energy(8000, 12000),
        "12k_16khz": band_energy(12000, 16000),
        "above_16khz": band_energy_above(16000)
    }

    # ============================================================
    # Brick-wall Detection
    # ============================================================
    diffs = np.diff(mean_db_per_bin)
    big_drop_idx = np.where(diffs < -20)[0]

    brick_wall = bool(big_drop_idx.size)
    brick_freq = float(freqs[big_drop_idx[0]]) if big_drop_idx.size else None

    # ============================================================
    # Spectral Notch Detection (Median-filtering)
    # ============================================================
    smooth = sps.medfilt(mean_db_per_bin, kernel_size=9)
    minima = sps.argrelextrema(smooth, np.less)[0]
    notches = []

    for m in minima:
        left = smooth[max(0, m - 6):m]
        right = smooth[m + 1:min(len(smooth), m + 7)]

        neighbor_peak = max(
            left.max() if left.size else -999,
            right.max() if right.size else -999
        )

        depth = neighbor_peak - smooth[m]

        if depth >= 15 and freqs[m] > 100:
            notches.append({
                "freq": float(freqs[m]),
                "depth_db": float(depth)
            })

    # ============================================================
    # Additional Spectral Descriptors
    # ============================================================
    centroid = float(np.mean(librosa.feature.spectral_centroid(S=S, sr=sr)))
    bandwidth = float(np.mean(librosa.feature.spectral_bandwidth(S=S, sr=sr)))
    flatness = float(np.mean(librosa.feature.spectral_flatness(S=S)))
    rolloff = float(np.mean(librosa.feature.spectral_rolloff(S=S, sr=sr)))


    return {
        "S_db": S_db,
        "freqs": freqs,
        "hop_length": hop_length,
        "n_fft": n_fft,

        "rolloff_85pct": freq_at_85,
        "rolloff_95pct": freq_at_95,
        "highest_freq_minus60db": highest_freq,

        "energy_distribution": energy_stats,

        "brick_wall_detected": brick_wall,
        "brick_wall_freq": brick_freq,

        "spectral_notches": notches,

        "spectral_centroid": centroid,
        "spectral_bandwidth": bandwidth,
        "spectral_flatness": flatness,
        "spectral_rolloff": rolloff,

        "hf_env": mean_db_per_bin,
        "lf_env": mean_db_per_bin[:200] if len(mean_db_per_bin) > 200 else mean_db_per_bin
    }