Update app.py

app.py

@@ -31,7 +31,10 @@ def predict_vad(audio, sr):
         start = (len(audio) - target_length) // 2
         audio = audio[start:start + target_length]
     
-    audio = audio / (np.max(np.abs(audio)) + 1e-8)
+    max_val = np.max(np.abs(audio))
+    if max_val > 0:
+        audio = audio / max_val
+    
     features = audio.reshape(1, target_length, 1)
     
     prediction = model.predict(features, verbose=0)
@@ -76,8 +79,8 @@ def analyze():
         else:
             y_mono = np.mean(y, axis=0)
         
-        n_channels = y.shape[0] if y.ndim > 1 else 1
-        duration = y.shape[-1] / sr
+        n_channels = int(y.shape[0]) if y.ndim > 1 else 1
+        duration = float(y.shape[-1] / sr)
         
         vad_label, vad_confidence, pred_class = predict_vad(y_mono, sr)
         
@@ -96,23 +99,46 @@ def analyze():
                 frame_label, frame_conf = vad_label, vad_confidence
             
             vad_timeline.append({
-                'start': round(start_sample / sr, 3),
-                'end': round(end_sample / sr, 3),
+                'start': float(round(start_sample / sr, 3)),
+                'end': float(round(end_sample / sr, 3)),
                 'label': frame_label,
-                'confidence': round(frame_conf, 3)
+                'confidence': float(round(frame_conf, 3))
             })
         
-        doa_angle = -25.5 + np.random.randn() * 10
-        doa_trajectory = [{'time': round(i * (duration/n_frames), 2), 'angle': round(doa_angle + np.random.randn()*3, 1)} for i in range(n_frames)]
+        doa_angle = float(-25.5 + np.random.randn() * 10)
+        doa_trajectory = [{'time': float(round(i * (duration/n_frames), 2)), 'angle': float(round(doa_angle + np.random.randn()*3, 1))} for i in range(n_frames)]
+        
+        voice_count = sum(1 for f in vad_timeline if f['label'] == 'voice')
+        voice_ratio = float(voice_count / len(vad_timeline))
         
-        voice_ratio = sum(1 for f in vad_timeline if f['label'] == 'voice') / len(vad_timeline)
+        mean_y = float(np.mean(y_mono**2))
+        snr = float(round(10 * np.log10(mean_y / 1e-10), 1)) if mean_y > 0 else 0.0
+        mean_doa = float(round(np.mean([d['angle'] for d in doa_trajectory]), 1))
         
         return jsonify({
             'success': True,
-            'metadata': {'channels': int(n_channels), 'sampleRate': int(sr), 'duration': round(duration, 2), 'samples': int(y.shape[-1])},
-            'vad': {'prediction': vad_label, 'confidence': round(vad_confidence, 3), 'predicted_class': pred_class, 'timeline': vad_timeline},
-            'doa': {'angle': round(doa_angle, 1), 'confidence': 0.89, 'trajectory': doa_trajectory},
-            'metrics': {'snr': round(10 * np.log10(np.mean(y_mono**2) / 1e-10), 1), 'voiceRatio': round(voice_ratio, 2), 'meanDoa': round(np.mean([d['angle'] for d in doa_trajectory]), 1)}
+            'metadata': {
+                'channels': n_channels,
+                'sampleRate': int(sr),
+                'duration': float(round(duration, 2)),
+                'samples': int(y.shape[-1])
+            },
+            'vad': {
+                'prediction': vad_label,
+                'confidence': float(round(vad_confidence, 3)),
+                'predicted_class': int(pred_class),
+                'timeline': vad_timeline
+            },
+            'doa': {
+                'angle': float(round(doa_angle, 1)),
+                'confidence': 0.89,
+                'trajectory': doa_trajectory
+            },
+            'metrics': {
+                'snr': snr,
+                'voiceRatio': float(round(voice_ratio, 2)),
+                'meanDoa': mean_doa
+            }
         })
     except Exception as e:
         import traceback