Data

Data Types

SKDH uses the following standardization of input data:

Name	Parameter Name	Description
Time	time	Time in seconds since 1970-01-01. These timestamps can either be naive or passed along-side time-zone information (see below).
Acceleration	accel	Acceleration normalized to gravity. Signal magnitude values at rest should be ~1.0
Angular Velocity	gyro	Angular velocity in radians per second.
Sampling Frequency	fs	Sampling frequency, in samples per second.

Time specification

Time information can be specified in two ways:

• naive/local timestamps: these timestamps have no timezone information associated with them. They are converted directly to datetimes (i.e. human readable) without any conversion. This will likely mean that if the recording duration crosses a time change (e.g. Daylight Savings Time) timestamps following the change will be shifted/offset by an hour.

• aware timestamps: these timestamps are “proper” UTC timestamps (seconds since 1970-01-01 UTC), but the time-zone name is also passed to any of the SKDH methods process (or SKDH pipeline run method) to provide the necessary information to convert these to local time.

IO methods usually produce naive/local timestamps if no timezone information (tz_name) is provided, as this information is not usually available from the files themselves. However, all IO methods should accept the tz_name parameter which allows the user to specify the timezone of the data, and convert the timestamps to aware timestamps.

Methods which produce final results/endpoints (e.g. GaitLumbar, Sleep, etc) will also accept the tz_name parameter. When passing tz_name to these (or any) methods, SKDH will assume that the timestamps passed in via time are UTC, and must be converted to the local time using the provided tz_name.

Data Flow

Data flow in SKDH pipelines is such that any parameters passed in are either used, potentially modified (e.g. accel in the CalibrateAccelerometer process), and then passed along to following processes, or simply passed along unused. This means that if you setup a pipeline, any additional parameters passed to the Pipeline.run method will be passed down to all processes in the pipeline.

For example, this is what allows a height value to be passed through a pipeline all the way to the GaitLumbar process where it is required to compute spatial gait metrics:

import skdh

pipeline = skdh.Pipeline()
pipeline.add(skdh.io.ReadCSV())
pipeline.add(skdh.preprocessing.CalibrateAccelerometer())
pipeline.add(skdh.gait.GaitLumbar())

pipeline.run(file='data.csv', height=1.8)  # height is passed all the way to GaitLumbar

This also means that if your data contains UTC timestamps (e.g. ActiGraph CSV files), passing the local timezone name will follow through the entire pipeline and make sure that the end results are in local time.