Update app.py

app.py

@@ -19,29 +19,21 @@ def smooth_signal(signal, window_len=11):
     smoothed = np.convolve(window/window.sum(), s, mode='valid')
     return smoothed
 
-def detect_beeps(audio_data, framerate, freq_zero, freq_one):
-    log_message("Starting beep detection.")
+def detect_beeps(audio_data, framerate, freq_zero, freq_one, window_size=2048, step_size=512):
+    log_message("Starting beep detection with sliding window.")
     
     binary_data = bytearray()
-    
-    # Calculate amplitude envelope to detect beep regions more clearly
-    envelope = np.abs(audio_data)
-    envelope = smooth_signal(envelope, window_len=100)
-    
-    # Detect peaks in the smoothed amplitude envelope
-    peaks, _ = find_peaks(envelope, height=0.05, distance=framerate // 10)
-    
+    num_windows = len(audio_data) // step_size
     previous_bit = None
-    for peak in peaks:
-        # Analyze a small window around each peak to capture the frequency
-        window_start = max(0, peak - framerate // 100)
-        window_end = min(len(audio_data), peak + framerate // 100)
-        window = audio_data[window_start:window_end]
+    
+    for i in range(0, len(audio_data) - window_size, step_size):
+        # Get the window of data
+        window = audio_data[i:i + window_size]
         
-        # Perform FFT to detect the dominant frequency
+        # Perform FFT to detect the dominant frequency in the window
         yf = fft(window)
-        xf = fftfreq(len(window), 1 / framerate)
-        magnitude = np.abs(yf[:len(window) // 2])
+        xf = fftfreq(window_size, 1 / framerate)
+        magnitude = np.abs(yf[:window_size // 2])
         dominant_freq = xf[np.argmax(magnitude)]
         
         # Classify based on the dominant frequency