Examples

Run a stand-alone process

import skdh

# load your data
time, accel = fn_to_get_data()

# setup and run the gait process. No prediction of gait periods, the entire
# recording/data is assumed to be gait data. Height is subject's height
# in meters
gait = skdh.gait.GaitLumbar()
gait_res = gait.predict(time=time, accel=accel, height=1.8)

Run a process with a prediction of gait periods

import skdh

# load your data
time, accel = fn_to_get_data()

# setup a pipeline to first predict gait periods and then estimate gait
# metrics during those periods.
pipeline = skdh.Pipeline()
pipeline.add(skdh.context.PredictGaitLumbarLgbm())
pipeline.add(skdh.gait.GaitLumbar())

# get the results
res = pipeline.run(time=time, accel=accel, height=1.8)
# gait results are in res['GaitLumbar']

Run a process starting with data ingestion

import skdh

# pipeline setup
pipeline = skdh.Pipeline()
pipeline.add(
    skdh.io.ReadCsv(
        time_col_name='Timestamp UTC',
        column_names={'accel': ['Accelerometer X', 'Accelerometer Y', 'Accelerometer Z']},
        drop_duplicate_timestamps=False,
        fill_gaps=True,
        fill_value={'accel': [0.0, 1.0, 0.0]},  # y axis is vertical
        to_datetime_kwargs={'unit': 's'},  # timestamp column is unix timestamps in seconds
        raw_conversions={'accel': 9.81},   # convert from m/s^2 to g
        read_csv_kwargs={'skiprows': 6},  # skip the first 6 rows - header information
    )
)
# find days (midnight-midnight) so that get gait results already with an associated
# day number & date
pipeline.add(skdh.preprocessing.GetDayWindowIndices(bases=[0], periods=[24]))
pipeline.add(skdh.preprocessing.CalibrateAccelerometer())  # calibrate the accelerometer
pipeline.add(skdh.context.PredictGaitLumbarLgbm())
pipeline.add(
    skdh.gait.GaitLumbar(),  # default parameters
    save_file="{file}_gait_results.csv"  # automatically save gait results to a file
    # this will use the CSV file name as the start of the output file name
)

# run the pipeline
res = pipeline.run(file='your_data.csv', height=1.8)

Follow a IPython notebook example

A notebook stepping through the process of setting up SKDH processes and then building up a pipeline can be found in the examples directory of the respository, along with a short sample data file. The notebook is called skdh_example.ipynb.