skdh.activity.SignalFeatures

class skdh.activity.SignalFeatures(window_minutes=15, window_skip_percentage=0.5, state='wake')

Compute various signal features on the raw acceleration metric.

Parameters:

window_minutesint, optional

Number of minutes for each window. Default is 15.

window_skip_percentagefloat, optional

Skip percentage of the window. Values between 0 and 1. Default is 0.5. Will be trimmed between 0.01 and 0.99.

state{‘wake’, ‘sleep’}

State during which the endpoint is being computed.

Methods

predict(results, i, accel_metric, ...)	Saves the signal features values.
reset_cached()	Called after all the blocks during the desired state have been run.

predict(results, i, accel_metric, accel_metric_60, epoch_s, epochs_per_min)

Saves the signal features values.

Parameters:

resultsdict

Dictionary containing the initialized results arrays. Keys in results are taken from the names of endpoints.

iint

Index of the day, used to index into individual result arrays, e.g. results[self.name][i] = 5.0

accel_metricnumpy.ndarray

Computed acceleration metric (e.g. ENMO).

accel_metric_60numpy.ndarray

Computed acceleration metric for a 60 second window.

epoch_sint

Duration in seconds of each sample of accel_metric.

epochs_per_minint

Number of epochs per minute.

reset_cached()

Called after all the blocks during the desired state have been run. Can be used to calculate results on all data for the day/state.