Update app.py

app.py

@@ -44,14 +44,16 @@ def predict_text_emotion(text):
 # Extract audio features and predict emotion
 def extract_audio_features(audio_data, sample_rate):
     if not isinstance(audio_data, np.ndarray):
-        audio_data = np.array(audio_data, dtype=np.float32)  # Ensure it is a NumPy array with float type
+        audio_data = np.array(audio_data, dtype=np.float32)  # Ensure it's a NumPy array with float type
     else:
         audio_data = audio_data.astype(np.float32)  # Convert to float32
 
-    mfcc = np.mean(librosa.feature.mfcc(y=audio_data, sr=sample_rate, n_mfcc=40).T, axis=0)
-    features = np.expand_dims(mfcc, axis=0)
+    mfcc = librosa.feature.mfcc(y=audio_data, sr=sample_rate, n_mfcc=704)
+    mfcc = np.mean(mfcc.T, axis=0)  # Compute mean across time
+    features = np.expand_dims(mfcc, axis=0)  # Add batch dimension
     return features
 
+
 def predict_audio_emotion(audio_data, sample_rate):
     features = extract_audio_features(audio_data, sample_rate)
     features = np.reshape(features, (1, 40))  # Match model expected input