fix opensmile

app/main.py

@@ -17,6 +17,7 @@ import pandas as pd
 import numpy as np
 import librosa.display
 import soundfile as sf
+import opensmile
 
 
 load_dotenv()
@@ -198,32 +199,39 @@ def preprocess_audio(path, save_dir):
     
     return 'success'
 
+smile = opensmile.Smile(
+    feature_set=opensmile.FeatureSet.ComParE_2016,
+    feature_level=opensmile.FeatureLevel.Functionals,
+)
 
 def extract_features(file_path):
-    # Load the audio file
-    y, sr = librosa.load(file_path, sr=None, dtype=np.float32)
+    # # Load the audio file
+    # y, sr = librosa.load(file_path, sr=None, dtype=np.float32)
+    
+    # # Extract MFCCs
+    # mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20)
+    # mfccs_mean = pd.Series(mfccs.mean(axis=1), index=[f'mfcc_{i}' for i in range(mfccs.shape[0])])
     
-    # Extract MFCCs
-    mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20)
-    mfccs_mean = pd.Series(mfccs.mean(axis=1), index=[f'mfcc_{i}' for i in range(mfccs.shape[0])])
+    # # Extract Spectral Features
+    # spectral_centroids = pd.Series(np.mean(librosa.feature.spectral_centroid(y=y, sr=sr)), index=['spectral_centroid'])
+    # spectral_rolloff = pd.Series(np.mean(librosa.feature.spectral_rolloff(y=y, sr=sr)), index=['spectral_rolloff'])
+    # spectral_flux = pd.Series(np.mean(librosa.onset.onset_strength(y=y, sr=sr)), index=['spectral_flux'])
+    # spectral_contrast = pd.Series(np.mean(librosa.feature.spectral_contrast(S=np.abs(librosa.stft(y)), sr=sr), axis=1), index=[f'spectral_contrast_{i}' for i in range(librosa.feature.spectral_contrast(S=np.abs(librosa.stft(y)), sr=sr).shape[0])])
     
-    # Extract Spectral Features
-    spectral_centroids = pd.Series(np.mean(librosa.feature.spectral_centroid(y=y, sr=sr)), index=['spectral_centroid'])
-    spectral_rolloff = pd.Series(np.mean(librosa.feature.spectral_rolloff(y=y, sr=sr)), index=['spectral_rolloff'])
-    spectral_flux = pd.Series(np.mean(librosa.onset.onset_strength(y=y, sr=sr)), index=['spectral_flux'])
-    spectral_contrast = pd.Series(np.mean(librosa.feature.spectral_contrast(S=np.abs(librosa.stft(y)), sr=sr), axis=1), index=[f'spectral_contrast_{i}' for i in range(librosa.feature.spectral_contrast(S=np.abs(librosa.stft(y)), sr=sr).shape[0])])
+    # # Extract Pitch
+    # pitches, magnitudes = librosa.piptrack(y=y, sr=sr)
+    # pitch_mean = pd.Series(np.mean(pitches[pitches != 0]), index=['pitch_mean'])  # Average only non-zero values
     
-    # Extract Pitch
-    pitches, magnitudes = librosa.piptrack(y=y, sr=sr)
-    pitch_mean = pd.Series(np.mean(pitches[pitches != 0]), index=['pitch_mean'])  # Average only non-zero values
+    # # Extract Zero Crossings
+    # zero_crossings = pd.Series(np.mean(librosa.feature.zero_crossing_rate(y)), index=['zero_crossings'])
     
-    # Extract Zero Crossings
-    zero_crossings = pd.Series(np.mean(librosa.feature.zero_crossing_rate(y)), index=['zero_crossings'])
+    # # Combine all features into a single Series
+    # features = pd.concat([mfccs_mean, spectral_centroids, spectral_rolloff, spectral_flux, spectral_contrast, pitch_mean, zero_crossings])
     
-    # Combine all features into a single Series
-    features = pd.concat([mfccs_mean, spectral_centroids, spectral_rolloff, spectral_flux, spectral_contrast, pitch_mean, zero_crossings])
+    features = smile.process_file(file_path)
+    all_data = pd.concat([features], ignore_index=True)
 
-    return pd.DataFrame([features])
+    return pd.DataFrame([all_data])
 
 
 @app.post("/mlp")