log_smooth#

PySeismoSoil.helper_signal_processing.log_smooth(signal: ndarray, win_len: int = 15, window: Literal['flat', 'hanning', 'hamming', 'bartlett', 'blackman'] = 'hanning', lin_space: bool = True, fmin: float | None = None, fmax: float | None = None, n_pts: int = None, fix_ends: bool = True, beta1: float = 0.9, beta2: float = 0.9) → ndarray[source]#

Smooth a frequency spectrum with constant window size in logarithmic space.

Parameters:

  • signal (np.ndarray) – The signal to be smoothed. Must be a 1D numpy array.

  • win_len (int) – The length of the convolution window.

  • window (Literal['flat', 'hanning', 'hamming', 'bartlett', 'blackman']) – The name of the window.

  • lin_space (bool) – Whether the points of the signal is uniformly spaced linearly. If False, the signal is treated as uniformaly spaced logarithmically.

  • fmin (float | None) – Minimum frequency (in Hz) that the signal is spaced within. Only effective when lin_space is True.

  • fmax (float | None) – Maximum frequency (in Hz) that the signal is spaced within. Only effective when lin_space is True.

  • n_pts (int) – The number of points of the logarithmically interpolated the signal. Only effective when lin_space is True.

  • fix_ends (bool) – Whether to fix the two ends of the smoothed signal, so that the “boundary effect” from convolution can be corrected. If True, the first and last n points will be adjusted using the exponentially weighted averaging method. (n is half of win_len.)

  • beta1 (float) – The “strength” of exponentially weighted averaging. For the head and the tail ends, respectively. Values should be within [0, 1].

  • beta2 (float) – The “strength” of exponentially weighted averaging. For the head and the tail ends, respectively. Values should be within [0, 1].

Returns:

smoothed_signal – The smoothed signal which has the same dimension as the original signal.

Return type:

np.ndarray

Raises:

ValueError – When the input values are not entirely valid