Update app.py

app.py

@@ -634,59 +634,69 @@ def analyze_audio_segments(y, sr, model, segment_duration=10):
     return segment_results
 
 
-# ========================================
-# HYBRID APNEA DETECTION - NEW!
-# ========================================
 # ========================================
 # HYBRID APNEA DETECTION - RELAXED VALIDATION
 # ========================================
-def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
+def count_precise_apneas(y, sr, segment_results, min_apnea=1, max_apnea=60):
     """
     HYBRID APPROACH - Detects apnea ONLY between snoring periods
-    RELAXED VALIDATION: More lenient gap silence threshold
-    
+    FIXED: Apnea is only valid when there is confirmed snoring IMMEDIATELY
+           before AND immediately after the gap (consecutive snoring periods).
+    Gap silence range: 1 to 60 seconds.
+
     Risk Levels:
     - 10-15 seconds: LOW
     - 15-20 seconds: MEDIUM
     - 20+ seconds: HIGH
-    
+
     Returns: (apnea_count, apnea_events_for_annotations)
     """
     print("\n🎯 APNEA DETECTION (Only between SNORING)")
     print("   Stage 1: Identifying confirmed snoring periods...")
-    
+
     # STAGE 1: Get all confirmed snoring segments from segment_results
+    # Build a full ordered list of all segments with their snoring status
+    # so we can check immediate neighbors
+    all_segments_ordered = sorted(segment_results, key=lambda s: s['start_time'])
+
     snoring_periods = []
-    
-    for seg in segment_results:
+    for seg in all_segments_ordered:
         if seg['is_snoring']:  # Only HIGH and MEDIUM confidence snores
             snoring_periods.append({
                 'start': seg['start_time'],
                 'end': seg['end_time'],
                 'confidence': seg['confidence_level']
             })
-    
+
     print(f"      Found {len(snoring_periods)} snoring periods")
-    
+
     if len(snoring_periods) < 2:
         print("      ⚠️ Need at least 2 snoring periods to detect apnea")
         return 0, []
-    
+
     # STAGE 2: Find gaps BETWEEN snoring periods
-    print("   Stage 2: Finding gaps between snoring periods...")
-    
+    # FIXED: Only consider CONSECUTIVE snoring periods (i and i+1 in the
+    # snoring_periods list). This guarantees that the segment immediately
+    # before the gap is snoring AND the segment immediately after is snoring.
+    print("   Stage 2: Finding gaps between consecutive snoring periods...")
+
     apnea_candidates = []
-    
+
     for i in range(len(snoring_periods) - 1):
         current_snore = snoring_periods[i]
         next_snore = snoring_periods[i + 1]
-        
+
         # Calculate gap between end of current snore and start of next snore
         gap_start = current_snore['end']
         gap_end = next_snore['start']
         gap_duration = gap_end - gap_start
-        
-        # Only consider gaps in apnea range
+
+        # FIXED: The gap must be > 0 (there must actually be a silent gap
+        # between two snoring segments that are adjacent in the snoring list)
+        if gap_duration <= 0:
+            continue
+
+        # FIXED: Only consider gaps in the 1–60 second range
         if min_apnea <= gap_duration <= max_apnea:
             apnea_candidates.append({
                 'start': gap_start,
@@ -696,45 +706,48 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                 'after_snore': next_snore
             })
             print(f"      Found gap: {gap_start:.1f}s → {gap_end:.1f}s (duration: {gap_duration:.1f}s)")
-    
+        else:
+            print(f"      Skipping gap: {gap_start:.1f}s → {gap_end:.1f}s "
+                  f"(duration: {gap_duration:.1f}s) - outside 1-60s range")
+
     print(f"      Found {len(apnea_candidates)} potential apnea candidates")
-    
+
     # STAGE 3: Validate with ML model - RELAXED THRESHOLDS
     validated_apneas = []
-    
+
     if len(apnea_candidates) > 0:
         if finetuned_snore_model is not None:
             print("   Stage 3: ML validation (RELAXED - confirming reduced activity in gap)...")
-            
+
             for idx, candidate in enumerate(apnea_candidates):
                 # Extract gap audio
                 gap_start_sample = int(candidate['start'] * sr)
                 gap_end_sample = int(candidate['end'] * sr)
                 gap_audio = y[gap_start_sample:gap_end_sample]
-                
+
                 try:
                     # Resample to 22050 for model
                     if sr != SAMPLE_RATE:
                         gap_audio = librosa.resample(gap_audio, orig_sr=sr, target_sr=SAMPLE_RATE)
-                    
+
                     # Validate that the gap has REDUCED snoring activity
                     gap_prob = predict_snoring_segment(gap_audio, SAMPLE_RATE, finetuned_snore_model) if len(gap_audio) > sr else 0
-                    
+
                     # RELAXED VALIDATION - Three levels of acceptance
-                    
+
                     # Level 1: Very silent gap (HIGH confidence apnea)
                     very_silent = gap_prob < 0.40
-                    
+
                     # Level 2: Moderately quiet gap (MEDIUM confidence apnea)
                     moderately_quiet = gap_prob < 0.65
-                    
+
                     # Level 3: Less snoring than before/after (LOW confidence apnea)
                     # Accept if gap has less activity than surrounding snoring
                     less_activity = gap_prob < 0.75
-                    
+
                     # Accept if ANY level matches
                     is_valid = very_silent or moderately_quiet or less_activity
-                    
+
                     if is_valid:
                         # Determine confidence based on gap silence
                         if very_silent:
@@ -743,7 +756,7 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                             validation_confidence = "MEDIUM"
                         else:
                             validation_confidence = "LOW"
-                        
+
                         # Determine risk level based on duration
                         duration = candidate['duration']
                         if duration < 15:
@@ -752,7 +765,7 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                             risk_level = "MEDIUM"
                         else:  # >= 20 seconds
                             risk_level = "HIGH"
-                        
+
                         validated_apneas.append({
                             'label': 'APNEA',
                             'start_sec': round(candidate['start'], 2),
@@ -771,10 +784,10 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                     else:
                         print(f"      ❌ REJECTED: {candidate['start']:.1f}s → {candidate['end']:.1f}s "
                               f"(Gap has too much snoring: {gap_prob:.2f})")
-                        
+
                 except Exception as e:
                     print(f"      ⚠️ Validation error for candidate {idx+1}: {e}")
-                    
+
                     # On error, accept gaps >= 12 seconds
                     if candidate['duration'] >= 12:
                         duration = candidate['duration']
@@ -784,7 +797,7 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                             risk_level = "MEDIUM"
                         else:
                             risk_level = "HIGH"
-                        
+
                         validated_apneas.append({
                             'label': 'APNEA',
                             'start_sec': round(candidate['start'], 2),
@@ -805,7 +818,7 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                     risk_level = "MEDIUM"
                 else:
                     risk_level = "HIGH"
-                
+
                 validated_apneas.append({
                     'label': 'APNEA',
                     'start_sec': round(candidate['start'], 2),
@@ -813,9 +826,9 @@ def count_precise_apneas(y, sr, segment_results, min_apnea=10, max_apnea=90):
                     'duration': round(candidate['duration'], 2),
                     'risk_level': risk_level
                 })
-    
+
     print(f"   ✅ Final: {len(validated_apneas)} validated apnea events\n")
-    
+
     return len(validated_apneas), validated_apneas
 
 
@@ -983,13 +996,14 @@ async def analyze_audio(file: UploadFile = File(...)):
                 
                 annotations.append(annotation)
         
-        # NEW: HYBRID APNEA DETECTION (replaces old segment-based method)
+        # NEW: HYBRID APNEA DETECTION
+        # FIXED: min_apnea=1, max_apnea=60 (1 to 60 second gaps only)
         apnea_count, apnea_events = count_precise_apneas(
             y_clean,
             sr,
             segment_results,
-            min_apnea=10,
-            max_apnea=90
+            min_apnea=1,
+            max_apnea=60
         )
         
         # Add apnea events to annotations