data.py 文件源码

def save_audio_file(filename, quantized_signal, quantization_steps=256, format="16bit_pcm", sampling_rate=48000):
    quantized_signal = quantized_signal.astype(float)
    normalized_signal = (quantized_signal / quantization_steps - 0.5) * 2.0

    # inv mu-law companding transformation (ITU-T, 1988)
    mu = quantization_steps - 1
    signals_1d = np.sign(normalized_signal) * ((1 + mu) ** np.absolute(normalized_signal)) / mu

    if format == "16bit_pcm":
        max = 1<<15
        type = np.int16
    elif format == "32bit_pcm":
        max = 1<<31
        type = np.int32
    elif format == "8bit_pcm":
        max = 1<<8 - 1
        type = np.uint8
    signals_1d *= max

    audio = signals_1d.reshape((-1, 1)).astype(type)
    audio = np.repeat(audio, 2, axis=1)
    wavfile.write(filename, sampling_rate, audio)

# convert signal to 1xW image