Spaces:
Sleeping
Sleeping
File size: 3,063 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 81 82 83 84 85 86 87 88 89 90 | """eGeMAPS feature extraction using OpenSMILE."""
import logging
import numpy as np
import opensmile
from src.features.schemas import EGeMAPSFeatures
logger = logging.getLogger(__name__)
class EGeMAPSExtractor:
"""Extract eGeMAPS features using OpenSMILE."""
def __init__(self, feature_set: str = "eGeMAPSv02", feature_level: str = "Functionals"):
"""
Initialize eGeMAPS extractor.
Args:
feature_set: Feature set name (eGeMAPSv02, GeMAPSv01b, etc.)
feature_level: Feature level (Functionals or LowLevelDescriptors)
"""
self.feature_set = feature_set
self.feature_level = feature_level
try:
self.smile = opensmile.Smile(
feature_set=opensmile.FeatureSet[feature_set],
feature_level=opensmile.FeatureLevel[feature_level],
)
logger.info(f"OpenSMILE initialized: {feature_set}/{feature_level}")
except Exception as e:
logger.warning(f"OpenSMILE initialization failed: {e}")
self.smile = None
def extract(self, waveform: np.ndarray, sr: int) -> EGeMAPSFeatures:
"""
Extract eGeMAPS features.
Args:
waveform: Audio waveform
sr: Sample rate
Returns:
EGeMAPSFeatures with 88-dimensional feature vector
"""
logger.debug("Extracting eGeMAPS features")
if self.smile is None:
logger.warning("OpenSMILE not available, using placeholder features")
return self._placeholder_features()
try:
# Process audio
features_df = self.smile.process_signal(waveform, sr)
# Extract feature values (first row, all columns)
features = features_df.iloc[0].values.astype(np.float32)
# Get feature names
feature_names = list(features_df.columns)
# Ensure 88 dimensions (eGeMAPSv02 Functionals)
if len(features) != 88:
logger.warning(
f"Expected 88 eGeMAPS features, got {len(features)}. Padding/truncating."
)
if len(features) < 88:
features = np.pad(features, (0, 88 - len(features)), mode="constant")
else:
features = features[:88]
logger.info(f"Extracted eGeMAPS features: {len(features)} dims")
return EGeMAPSFeatures(features=features, feature_names=feature_names)
except Exception as e:
logger.error(f"eGeMAPS extraction failed: {e}")
return self._placeholder_features()
def _placeholder_features(self) -> EGeMAPSFeatures:
"""Return placeholder eGeMAPS features."""
# Return zeros as placeholder
features = np.zeros(88, dtype=np.float32)
feature_names = [f"egemaps_{i}" for i in range(88)]
logger.warning("Using placeholder eGeMAPS features (zeros)")
return EGeMAPSFeatures(features=features, feature_names=feature_names)
|