Delete seuilACTfiltre.py

seuilACTfiltre.py

-import sounddevice as sd
-import numpy as np
-import soundfile as sf
-from collections import deque
-from datetime import datetime
-from scipy.signal import butter, lfilter
-
-# --- Paramètres audio ---
-RATE = 44100
-CHANNELS = 2
-THRESHOLD = -50.0  # Ajuste ce seuil selon tes tests
-duration_before = 5  # secondes avant événement
-duration_after = 5   # secondes après l'événement
-
-# --- Buffers ---
-buffer = deque(maxlen=int(RATE * duration_before))
-intensity_buffer = deque(maxlen=int(RATE * 5))  # Buffer intensité pour 5 secondes
-
-# --- Filtrage passe-haut Butterworth ---
-def butter_highpass(cutoff, fs, order=4):
-    nyq = 0.5 * fs
-    normal_cutoff = cutoff / nyq
-    return butter(order, normal_cutoff, btype='high', analog=False)
-
-def highpass_filter(data, cutoff=1500.0, fs=44100, order=4):
-    b, a = butter_highpass(cutoff, fs, order=order)
-    return lfilter(b, a, data)
-
-# --- Fonction pour sauvegarder un fichier FLAC ---
-def save_wav(filename, data, samplerate):
-    filename = filename.replace('.wav', '.flac')
-    sf.write(filename, data, samplerate, format='FLAC')
-    print(f"Audio enregistré sous {filename}")
-
-# --- Fonction score ---
-def score(x, threshold, axis=0):
-    x = np.asarray(x)
-    x = x >= threshold
-    count = sum(x, axis)
-    s = sum(x, axis) / x.shape[axis]
-    return s, count
-
-# --- Activité acoustique ---
-def acoustic_activity(xdB, dB_threshold, axis=1):
-    ACTfract, ACTcount = score(xdB, dB_threshold, axis=axis)
-    ACTfract = ACTfract.tolist()
-    ACTcount = ACTcount.tolist()
-    ACTmean = dB2linear(xdB[xdB > dB_threshold], mode='power')
-    return ACTfract, ACTcount, ACTmean
-
-# --- Conversion dB vers linéaire ---
-def dB2linear(x, mode='amplitude', db_gain=0):
-    if mode == 'amplitude':
-        y = 10**((x - db_gain) / 20)
-    elif mode == 'power':
-        y = 10**((x - db_gain) / 10)
-    return y
-
-# --- Callback pour audio continu ---
-def audio_callback(indata, frames, time, status):
-    global buffer, intensity_buffer
-    buffer.extend(indata[:, 0])
-    intensity_buffer.extend(indata[:, 0])
-
-# --- Lancement du stream audio ---
-with sd.InputStream(samplerate=RATE, channels=1, callback=audio_callback, blocksize=1024):
-    print("Début de l'écoute en continu. Appuyez sur Ctrl+C pour arrêter.")
-    try:
-        while True:
-            sd.sleep(1000)  # actualisation chaque seconde
-
-            # --- Analyse d'intensité ---
-            current_window = highpass_filter(np.array(intensity_buffer), cutoff=1500.0, fs=RATE)
-
-            # --- Calcul de l'activité acoustique ---
-            xdB = 10 * np.log10(np.abs(current_window) ** 2 + 1e-11)
-            _, ACT, _ = acoustic_activity(xdB, dB_threshold=THRESHOLD + 12, axis=-1)
-            ACT = np.sum(np.asarray(ACT)) / RATE
-            print(f"Activité acoustique: {ACT:.2f}")
-
-            intensity_buffer.clear()
-
-            if ACT > 0.6:  # Seuil ajustable
-                print("Son détecté, enregistrement en cours...")
-
-                # --- Pré-enregistrement ---
-                previous_audio = np.array(buffer)
-
-                # --- Enregistrement post-détection ---
-                next_audio = sd.rec(int(RATE * duration_after), samplerate=RATE, channels=1)
-                sd.wait()
-
-                combined_audio = np.concatenate((previous_audio, next_audio[:, 0]))
-
-                # --- Filtrage de l'audio complet ---
-                filtered_audio = highpass_filter(combined_audio, cutoff=1500.0, fs=RATE)
-
-                # --- Sauvegarde ---
-                filename = datetime.now().strftime("record_%Y%m%d_%H%M%S.wav")
-                save_wav(filename, filtered_audio, RATE)
-
-                buffer.clear()
-                intensity_buffer.clear()
-                print("Écoute relancée...")
-
-    except KeyboardInterrupt:
-        print("Arrêt de l'écoute.")
-
-