phaseret_pghi_sample.py

import numpy as np
import soundfile as sf
from oct2py import octave
from scipy import signal

WIN_LEN = 512
HOP_LEN = 128
FFT_LEN = 512
WINDOW = "hann"
IN_WAV = "ONOMATOPEE300_001.wav"
OUT_WAV = "hoge.wav"


def main():
    """Reconstruct phase by using PGHI."""

    # 事前にoctave上で
    # ltfatstart & ltfatmex
    # phaseretstart & phaseretmex
    # を済ませておく
    octave.addpath(octave.genpath("/work/tamamori/ltfat-main"))
    octave.ltfatstart(0)  # 引数の0を省略するとoct2pyのエラーでこける
    octave.phaseretstart(0)  # 引数の0を省略するとoct2pyのエラーでこける

    # 振幅スペクトルの取得
    audio, fs = sf.read(IN_WAV)
    stfft = signal.ShortTimeFFT(
        win=signal.get_window(WINDOW, WIN_LEN), hop=HOP_LEN, fs=int(fs), mfft=FFT_LEN
    )
    amp_spec = np.abs(stfft.stft(audio))

    # PGHIに基づく位相復元
    gamma = octave.pghi_findgamma("hann", HOP_LEN, WIN_LEN)
    reconst_spec = octave.pghi(amp_spec, gamma, HOP_LEN, WIN_LEN)
    audio = stfft.istft(reconst_spec)
    sf.write(OUT_WAV, audio, fs)


if __name__ == "__main__":
    main()