Add YAMNet sound classification to headmic

New sound_id.py module with SoundClassifier class that runs YAMNet
(521 audio event categories) on CPU TFLite. Classifies audio every
0.5s from a ring buffer fed by the existing audio stream.

Categories: speech, alert, music, animal, household, environment, silence.
Smoothing via 20-sample history window for stable dominant category.

New endpoints: GET /sounds, GET /sounds/history
Updated: /health (sound_classification_enabled), /status (audio_scene)
Graceful degradation if model files not present.

Model download (not tracked in git):
  curl -sL 'https://tfhub.dev/google/lite-model/yamnet/classification/tflite/1?lite-format=tflite' -o models/yamnet.tflite

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

headmic.py

@@ -34,6 +34,7 @@ import wave
 from pathlib import Path
 from typing import Optional, List
 
+import numpy as np
 import httpx
 import pvporcupine
 import webrtcvad
@@ -117,9 +118,15 @@ class ServiceState:
         self.last_wake_time: Optional[float] = None
         self.wake_count = 0
         self.error: Optional[str] = None
+        self.audio_scene: Optional[dict] = None
+        self.sound_classification_enabled: bool = False
 
 state = ServiceState()
 
+# Sound classifier globals
+sound_classifier = None
+sound_ring_buffer = None  # collections.deque, filled by listener_loop
+
 
 # ============================================================================
 # Audio Stream using ALSA directly (arecord)
@@ -254,7 +261,11 @@ def listener_loop():
             
             # Convert bytes to int16 array for Porcupine
             pcm = struct.unpack_from("h" * 512, frame_data)
-            
+
+            # Feed sound classifier ring buffer
+            if sound_ring_buffer is not None:
+                sound_ring_buffer.append(frame_data)
+
             # Check for wake word
             keyword_index = porcupine.process(pcm)
             
@@ -327,6 +338,31 @@ def listener_loop():
         logger.info("Listener stopped")
 
 
+# ============================================================================
+# Sound Classification Thread
+# ============================================================================
+
+def sound_classifier_loop():
+    """Background thread for continuous sound classification."""
+    global state
+    logger.info("Sound classifier thread started")
+    while state.running:
+        if sound_ring_buffer is None or len(sound_ring_buffer) < 30:
+            time.sleep(0.1)
+            continue
+
+        try:
+            frames = list(sound_ring_buffer)
+            audio = np.frombuffer(b"".join(frames), dtype=np.int16)
+            result = sound_classifier.classify(audio)
+            state.audio_scene = result
+        except Exception as e:
+            logger.warning("Sound classification error: %s", e)
+
+        time.sleep(0.5)
+    logger.info("Sound classifier thread stopped")
+
+
 # ============================================================================
 # FastAPI
 # ============================================================================
@@ -336,7 +372,30 @@ app = FastAPI(title="HeadMic", description="Vixy's Ears 🦊👂")
 
 @app.on_event("startup")
 async def startup():
+    global sound_classifier, sound_ring_buffer
+
     state.running = True
+
+    # Init sound classifier (optional — graceful if model missing)
+    model_dir = Path(__file__).parent / "models"
+    model_path = model_dir / "yamnet.tflite"
+    class_map_path = model_dir / "yamnet_class_map.csv"
+    if model_path.exists() and class_map_path.exists():
+        try:
+            from sound_id import SoundClassifier
+            sound_classifier = SoundClassifier(str(model_path), str(class_map_path))
+            # 31 frames of 512 samples = ~0.99s at 16kHz
+            sound_ring_buffer = collections.deque(maxlen=31)
+            state.sound_classification_enabled = True
+            logger.info("Sound classification enabled (YAMNet)")
+
+            sc_thread = threading.Thread(target=sound_classifier_loop, daemon=True)
+            sc_thread.start()
+        except Exception as e:
+            logger.warning("Sound classification unavailable: %s", e)
+    else:
+        logger.info("Sound classification models not found, skipping")
+
     thread = threading.Thread(target=listener_loop, daemon=True)
     thread.start()
     logger.info("HeadMic started")
@@ -365,6 +424,7 @@ async def health():
         "recording": state.recording,
         "processing": state.processing,
         "wake_count": state.wake_count,
+        "sound_classification_enabled": state.sound_classification_enabled,
         "error": state.error
     }
 
@@ -378,6 +438,7 @@ async def status():
         "last_transcription": state.last_transcription,
         "last_wake_time": state.last_wake_time,
         "wake_count": state.wake_count,
+        "audio_scene": state.audio_scene["dominant_category"] if state.audio_scene else None,
         "error": state.error
     }
 
@@ -390,6 +451,26 @@ async def last():
     }
 
 
+@app.get("/sounds")
+async def sounds():
+    """Current audio scene classification."""
+    if not state.sound_classification_enabled:
+        raise HTTPException(status_code=503, detail="Sound classification not available")
+    if state.audio_scene is None:
+        return {"category": None, "top_classes": [], "dominant_category": None, "timestamp": None}
+    return state.audio_scene
+
+
+@app.get("/sounds/history")
+async def sounds_history(seconds: int = 30):
+    """Recent sound classification history."""
+    if not state.sound_classification_enabled:
+        raise HTTPException(status_code=503, detail="Sound classification not available")
+    if sound_classifier is None:
+        return {"history": []}
+    return {"history": sound_classifier.get_history(seconds)}
+
+
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run(app, host="0.0.0.0", port=8446)