| import librosa | |
| import numpy as np | |
| import soundfile as sf | |
| def read_audio(audio_path, fs=16000): | |
| if audio_path.endswith('.wav'): | |
| # X, fs_orig = sf.read(audio_path) | |
| X, fs_orig = librosa.load(audio_path, sr=None) | |
| if fs_orig != fs: | |
| X = librosa.resample(X, orig_sr=fs_orig, target_sr=fs) | |
| else: | |
| X = np.load(audio_path) | |
| fs = 16000 | |
| return X, fs |