# yasa.stft_power

yasa.stft_power(data, sf, window=2, step=0.2, band=(1, 30), interp=True, norm=False)[source]

Compute the pointwise power via STFT and interpolation.

Parameters
dataarray_like

Single-channel data.

sffloat

Sampling frequency of the data.

windowint

Window size in seconds for STFT. 2 or 4 seconds are usually a good default. Higher values = higher frequency resolution = lower time resolution.

stepint

Step in seconds for the STFT. A step of 0.2 second (200 ms) is usually a good default.

• If step == 0, overlap at every sample (slowest)

• If step == nperseg, no overlap (fastest)

Higher values = higher precision = slower computation.

bandtuple or None

Broad band frequency range. Default is 1 to 30 Hz.

interpboolean

If True, a cubic interpolation is performed to ensure that the output is the same size as the input (= pointwise power).

normbool

If True, return bandwise normalized band power, i.e. for each time point, the sum of power in all the frequency bins equals 1.

Returns
fnumpy.ndarray

Frequency vector

tnumpy.ndarray

Time vector

Sxxnumpy.ndarray

Power in the specified frequency bins of shape (f, t)

Notes

2D Interpolation is done using scipy.interpolate.RectBivariateSpline which is much faster than scipy.interpolate.interp2d for a rectangular grid. The default is to use a bivariate spline with 3 degrees.