Spatial Parameter Estimation#

This example illustrates how Random Forest by the SpatialParamsRandomForest can be used to estimate spatial gait parameters, including step_length and stride_length.

The gait event detection algorithm DiaoAdaptedEventDetection is applied to obtain gait events.

Getting some example data#

For this we take some example data that contains regular walking movements.

import pandas as pd

from eargait import EarGait
from eargait.event_detection import DiaoAdaptedEventDetection
from eargait.preprocessing import align_gravity_and_convert_ear_to_ebf, convert_ear_to_ebf, load
from eargait.spatial_params import SpatialParamsRandomForest
from eargait.utils.example_data import get_mat_example_data_path

# data directory
data_path = get_mat_example_data_path()

Loading the data#

A data session refers to a recording by signia hearing aids. A session can consist of a single *.mat file or two *.mat file, for left and right ear, respectively. The session is loaded using the local path data_path` of the directory, in which the matlab file(s) are stored. For more options regardind loading the data see Load Data by Signia Hearing Aids. example_gait_event_detection

target_sample_rate = 50
session = load(data_path, target_sample_rate_hz=target_sample_rate, skip_calibration=False)
session.info

/home/docs/checkouts/readthedocs.org/user_builds/eargait/envs/latest/lib/python3.9/site-packages/nilspodlib/calibration_utils.py:190: CalibrationWarning: For the sensor 001 no calibration could be located that was in 60 days, 0:00:00 of the 2021-06-01 10:53:58+00:00.The closest calibration is 222 days, 23:54:02 away.
  return find_closest_calibration_info_to_date(
/home/docs/checkouts/readthedocs.org/user_builds/eargait/envs/latest/lib/python3.9/site-packages/nilspodlib/calibration_utils.py:190: CalibrationWarning: For the sensor 002 no calibration could be located that was in 60 days, 0:00:00 of the 2021-06-01 10:53:56+00:00.The closest calibration is 222 days, 23:54:04 away.
  return find_closest_calibration_info_to_date(

<signialib.header._ProxyHeader object at 0x7ff6463bb730>

Gravity alignment and data transformation into body frame#

Align session to gravity and transform coordinate system into body frame

ear_data = align_gravity_and_convert_ear_to_ebf(session)

# Alternatively, you can skip the gravity alignment by using the following function:  convert_ear_to_ebf
# ear_data = convert_ear_to_ebf(session)

/home/docs/checkouts/readthedocs.org/user_builds/eargait/checkouts/latest/eargait/preprocessing/rotations.py:170: UserWarning: No gravity alignment method provided. Using MeanTrimGravityAlignment.
  warnings.warn("No gravity alignment method provided. Using MeanTrimGravityAlignment.")

Extract walking interval#

Note: Here prior knowledge about walking sequence within the given data session are applied. load csv file containing walking bouts

rescale_factor = 200 / target_sample_rate
walking_bout_list = pd.read_csv(data_path.parent.joinpath("walking_bout_indices.csv"))
interval = (int(walking_bout_list["start"][0] / rescale_factor), int(walking_bout_list["stop"][0] / rescale_factor))

# alternative if walking bout indices are already known, e.g.
# interval = (14317, 17637)    --> indices need to be replaced based on data

ear_data_short = {}
for side in ear_data.keys():
    ear_data_short[side] = ear_data[side][interval[0] : interval[1]]

Initializing event detection algorithm#

Recommended parameters: apply filter = True <br /> sampling_rate_hz needs to correspond to target_sample_rate_hz <br /> window_length should be equal to sampling_rate_hz

event_detection_algorithm = DiaoAdaptedEventDetection(
    sample_rate_hz=target_sample_rate, window_length=target_sample_rate
)

Initializing spatial parameter estimation method#

Two Alternatives 1. spatial_method = SpatialParamsRandomForestDemographics(target_sample_rate, age, gender, height, weight) 2. spatial_method = SpatialParamsCNN(target_sample_rate)

spatial_method = SpatialParamsRandomForest(target_sample_rate)

Initializing Gait Analysis Pipeline#

Recommended parameters: sampling_rate_hz needs to correspond to target_sample_rate_hz <br />

ear_gait = EarGait(
    sample_rate_hz=target_sample_rate,
    event_detection_method=event_detection_algorithm,
    spatial_params_method=spatial_method,
    bool_use_event_list_consistent=True,
)

Detect gait events of gait sequence#

ear_gait.detect(ear_data_short)
gait_events = ear_gait.event_list

Get spatial parameter for walking bout#

spatial_params = ear_gait.spatial_params
spatial_params

{'right_sensor':       step_length  stride_length  gait_velocity           side
s_id
           NaN            NaN            NaN  contralateral
      0.694539       1.389077       1.389077    ipsilateral
      0.763567       1.527135       1.527135  contralateral
      0.836603       1.673206       1.673206    ipsilateral
      0.833778       1.667556       1.737037  contralateral
      0.835001       1.670002       1.739585    ipsilateral
      0.860160       1.720320       1.720320  contralateral
      0.830181       1.660363       1.729544    ipsilateral
      0.820921       1.641841       1.710251  contralateral
      0.794940       1.589879       1.589879    ipsilateral
     0.630958       1.261916       1.261916  contralateral
     0.605793       1.211586       1.262069    ipsilateral
     0.612922       1.225844       1.532306  contralateral
     0.682229       1.364458       1.421311    ipsilateral
     0.783562       1.567124       1.567124  contralateral
     0.837847       1.675695       1.675695    ipsilateral
     0.824688       1.649375       1.718099  contralateral
     0.859322       1.718644       1.718644    ipsilateral
     0.834872       1.669745       1.739317  contralateral
     0.856414       1.712828       1.712828    ipsilateral
     0.825928       1.651856       1.651856  contralateral
     0.781277       1.562554       1.627661    ipsilateral
     0.698586            NaN       1.397171  contralateral, 'left_sensor':       step_length  stride_length  gait_velocity           side
s_id
           NaN            NaN            NaN    ipsilateral
      0.688018       1.376036       1.376036  contralateral
      0.778775       1.557550       1.557550    ipsilateral
      0.831000       1.662000       1.662000  contralateral
      0.836897       1.673794       1.743536    ipsilateral
      0.833923       1.667845       1.737339  contralateral
      0.843920       1.687840       1.758166    ipsilateral
      0.852531       1.705063       1.705063  contralateral
      0.821491       1.642982       1.711439    ipsilateral
      0.794771       1.589542       1.655773  contralateral
          NaN            NaN            NaN    ipsilateral
          NaN            NaN            NaN    ipsilateral
          NaN            NaN            NaN  contralateral
     0.802355       1.604711       1.604711    ipsilateral
     0.829709       1.659418       1.728561  contralateral
     0.844934       1.689869       1.689869    ipsilateral
     0.840688       1.681375       1.751433  contralateral
     0.850179       1.700358       1.700358    ipsilateral
     0.839817       1.679635       1.679635  contralateral
     0.827547       1.655094       1.724056    ipsilateral
     0.798630       1.597260       1.597260  contralateral
     0.696382            NaN       1.392764    ipsilateral}

Get average spatial parameter over walking bout#

spatial_params_average = ear_gait.average_spatial_params
spatial_params_average

{'right_sensor':       step_length  stride_length  gait_velocity
mean     0.777459       1.562429       1.595547
std      0.083571       0.167422       0.156308, 'left_sensor':       step_length  stride_length  gait_velocity
mean     0.811754       1.637081       1.654197
std      0.048092       0.079411       0.112798}

Total running time of the script: ( 0 minutes 5.744 seconds)

Estimated memory usage: 12 MB

Gallery generated by Sphinx-Gallery