analyzer.py 文件源码

def extract(filename, fft_size=FFT_SIZE, dtype=np.float32):
    ''' Basic (WORLD) feature extraction ''' 
    x, _ = librosa.load(filename, sr=args.fs, mono=True, dtype=np.float64)
    features = wav2pw(x, args.fs, fft_size=fft_size)
    ap = features['ap']
    f0 = features['f0'].reshape([-1, 1])
    sp = features['sp']
    en = np.sum(sp + EPSILON, axis=1, keepdims=True)
    sp = np.log10(sp / en)
    return np.concatenate([sp, ap, f0, en], axis=1).astype(dtype)