How to use the yasa.spectral.stft_power function in yasa

mask = np.ones(data.size, dtype=bool)

    # Bandpass filter
    data = filter_data(data, sf, freq_broad[0], freq_broad[1], method='fir',
                       verbose=0)

    # The width of the transition band is set to 1.5 Hz on each side,
    # meaning that for freq_sp = (12, 15 Hz), the -6 dB points are located at
    # 11.25 and 15.75 Hz.
    data_sigma = filter_data(data, sf, freq_sp[0], freq_sp[1],
                             l_trans_bandwidth=1.5, h_trans_bandwidth=1.5,
                             method='fir', verbose=0)

    # Compute the pointwise relative power using interpolated STFT
    # Here we use a step of 200 ms to speed up the computation.
    f, t, Sxx = stft_power(data, sf, window=2, step=.2, band=freq_broad,
                           interp=False, norm=True)
    idx_sigma = np.logical_and(f >= freq_sp[0], f <= freq_sp[1])
    rel_pow = Sxx[idx_sigma].sum(0)

    # Let's interpolate `rel_pow` to get one value per sample
    # Note that we could also have use the `interp=True` in the `stft_power`
    # function, however 2D interpolation is much slower than
    # 1D interpolation.
    func = interp1d(t, rel_pow, kind='cubic', bounds_error=False,
                    fill_value=0)
    t = np.arange(data.size) / sf
    rel_pow = func(t)

    # Now we apply moving RMS and correlation on the sigma-filtered signal
    _, mcorr = moving_transform(data_sigma, data, sf, window=.3, step=.1,
                                method='corr', interp=True)