Spaces:
Sleeping
Sleeping
File size: 2,068 Bytes
4e9a3bc | 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 | """MFCC feature extraction."""
import logging
import numpy as np
import librosa
from src.features.schemas import MFCCFeatures
logger = logging.getLogger(__name__)
class MFCCExtractor:
"""Extract MFCC features with delta and delta-delta."""
def __init__(
self,
n_mfcc: int = 13,
n_fft: int = 2048,
hop_length: int = 512,
n_mels: int = 128,
fmin: float = 0,
fmax: float = 8000,
delta_width: int = 9,
):
"""Initialize MFCC extractor."""
self.n_mfcc = n_mfcc
self.n_fft = n_fft
self.hop_length = hop_length
self.n_mels = n_mels
self.fmin = fmin
self.fmax = fmax
self.delta_width = delta_width
def extract(self, waveform: np.ndarray, sr: int) -> MFCCFeatures:
"""
Extract MFCC features.
Args:
waveform: Audio waveform
sr: Sample rate
Returns:
MFCCFeatures with 39-dimensional feature vectors
"""
logger.debug(f"Extracting MFCCs: n_mfcc={self.n_mfcc}")
# Extract MFCCs
mfcc = librosa.feature.mfcc(
y=waveform,
sr=sr,
n_mfcc=self.n_mfcc,
n_fft=self.n_fft,
hop_length=self.hop_length,
n_mels=self.n_mels,
fmin=self.fmin,
fmax=self.fmax,
)
# Compute deltas
delta = librosa.feature.delta(mfcc, width=self.delta_width)
delta_delta = librosa.feature.delta(mfcc, order=2, width=self.delta_width)
# Combine: (13, time) + (13, time) + (13, time) = (39, time)
combined = np.vstack([mfcc, delta, delta_delta])
# Compute statistics
mean = np.mean(combined, axis=1)
std = np.std(combined, axis=1)
logger.info(f"Extracted MFCCs: shape={combined.shape}")
return MFCCFeatures(
mfcc=mfcc,
delta=delta,
delta_delta=delta_delta,
combined=combined,
mean=mean,
std=std,
)
|