def load_audio(audio_path, del_silence): sound = np.memmap(audio_path, dtype='h', mode='r').astype('float32') if del_silence: non_silence_indices = split(sound, top_db=30) sound = np.concatenate([sound[start:end] for start, end in non_silence_indices]) sound /= 32767 # normalize audio return sound
def extract_noise(self, audio_path): try: signal = np.memmap(audio_path, dtype='h', mode='r').astype('float32') non_silence_indices = split(signal, top_db=30) for (start, end) in non_silence_indices: signal[start:end] = 0 noise = signal[signal != 0] return noise / 32767 except RuntimeError: logger.info("RuntimeError in {0}".format(audio_path)) return None except ValueError: logger.info("RuntimeError in {0}".format(audio_path)) return None