osl_dynamics.data.processing#

Functions to process data.

Functions#

standardize(x[, axis, create_copy])

Standardizes a time series.

time_embed(x, n_embeddings)

Performs time embedding.

temporal_filter(x, low_freq, high_freq, sampling_frequency)

Applies temporal filtering.

amplitude_envelope(x)

Calculates amplitude envelope.

moving_average(x, n_window)

Calculates a moving average over a sliding window along the time axis.

downsample(x, new_freq, old_freq)

Downsample.

Module Contents#

osl_dynamics.data.processing.standardize(x, axis=0, create_copy=True)[source]#

Standardizes a time series.

Returns a time series with zero mean and unit variance.

Parameters:

  • x (np.ndarray) – Time series data. Shape must be (n_samples, n_channels).

  • axis (int, optional) – Axis on which to perform the transformation.

  • create_copy (bool, optional) – Should we return a new array containing the standardized data or modify the original time series array?

Returns:

X – Standardized data.

Return type:

np.ndarray

osl_dynamics.data.processing.time_embed(x, n_embeddings)[source]#

Performs time embedding.

Parameters:

  • x (np.ndarray) – Time series data. Shape must be (n_samples, n_channels).

  • n_embeddings (int) – Number of samples in which to shift the data.

Returns:

X – Time embedded data. Shape is (n_samples - n_embeddings + 1, n_channels * n_embeddings).

Return type:

sliding_window_view

osl_dynamics.data.processing.temporal_filter(x, low_freq, high_freq, sampling_frequency, order=5)[source]#

Applies temporal filtering.

Parameters:

  • x (np.ndarray) – Time series data. Shape must be (n_samples, n_channels).

  • low_freq (float) – Frequency in Hz for a high pass filter.

  • high_freq (float) – Frequency in Hz for a low pass filter.

  • sampling_frequency (float) – Sampling frequency in Hz.

  • order (int, optional) – Order for a butterworth filter.

Returns:

X – Filtered time series. Shape is (n_samples, n_channels).

Return type:

np.ndarray

osl_dynamics.data.processing.amplitude_envelope(x)[source]#

Calculates amplitude envelope.

Parameters:

x (np.ndarray) – Time series data. Shape must be (n_samples, n_channels).

Returns:

X – Amplitude envelope data. Shape is (n_samples, n_channels).

Return type:

np.ndarray

osl_dynamics.data.processing.moving_average(x, n_window)[source]#

Calculates a moving average over a sliding window along the time axis.

This function uses a cumulative-sum trick for efficiency and returns only resulting values where the full window fits.

Parameters:

  • x (np.ndarray) – Time series data. Shape must be (n_samples, n_channels).

  • n_window (int) – Number of data points in the sliding window. Must be odd.

Returns:

X – Time series with sliding window applied. Shape is (n_samples - n_window + 1, n_channels).

Return type:

np.ndarray

osl_dynamics.data.processing.downsample(x, new_freq, old_freq)[source]#

Downsample.

Parameters:

  • x (np.ndarray) – Time series data. Shape must be (n_samples, n_channels).

  • old_freq (float) – Old sampling frequency in Hz.

  • new_freq (float) – New sampling frequency in Hz.

Returns:

X – Downsampled time series. Shape is (n_samples * new_freq / old_freq, n_channels).

Return type:

np.ndarray