# # Physical activity

The algorithms in this chapter are useful to obtain information about the physical activity of a participant (intensity of movements, steps, activity class, etc.). This infomation can then be further used to estimate the energy expenditure.

## # Movement Acceleration (MovementAcceleration)

Movement Acceleration (aka Movement Acceleration Intensity, MAI) is a physical activity metric that outputs values that have a very good correlation to the intensity of bodily movements. For daily activities good sensor positions are at the participant's center of gravity (hip, chest). For each output interval (default is 1min) the values aggregated by calculating the mean value of all samples in the interval. The unit of the output parameter Movement Acceleration Intensity is [g], i.e. multiples of earth gravity (1g = 9,81 m/s2).

The algorithm uses raw 3D acceleration signals from the acceleration sensor. The signals from each axis is bandpass filtered (Butterworth filter 0.25-11Hz, order 4) to remove the offset due to the acceleration of gravity and to remove high frequency content that does not directly relate to movements. From the resulting signals for each sample the vector magnitude is calculated:

$ma_i = \sqrt{ax_{bp}^2 + ay_{bp}^2 + az_{bp}^2}$

For each output interval (default is 1min) the vector magnitude is aggregated by calculating the mean value of all samples in the interval. The unit of the output parameter Movement Acceleration Intensity is [g], i.e. multiples of earth gravity (1g = 9,81 m/s2). Also have a look at the physical activity metrics chapter.

movisensLIVE: Movement acceleration can be calculated live on the sensor.

## # Wear detection (NonWearTime)

This uses the acceleration data to reveal the times when the participant has removed the sensor. Internally wear time is calculated in 30s intervals.

Sensor state Coding
Worn 0
Not worn 1

## # Acceleration along the body axes (AccDown, AccForward, AccRight)

This parameter displays the acceleration at the sensor position along the three axes of the participant’s body. These axes are defined in reference to an upright standing position, arms hanging down, palms pointing inwards.

Signal name Direction of axis
AccDown Acceleration down
AccForward Acceleration forward
AccRight Acceleration to the right

Please check that the sensor position has been set correctly. The parameter has a fixed output that’s based on the sample rate of the raw acceleration data (default is 64Hz).

## # Inclination (InclinsationDown, InclinationForward, InclinationRight)

This parameter displays the inclination of the body axes at the sensor location against the vertical. It calculates the mean inclinations of the three body axes from the acceleration signal and displays the value for each inclination in degrees. The values range from 0° to 180°. Internally the inclination calculated in intervals of 4s.

Signal name Definition of inclination
InclinationDown Inclination of the body axis „down“ against the vertical
InclinationForward Inclination of the body axis „forward“ against the vertical
InclinationRight Inclination of the body axis „to the right“ against the vertical

Please check that the sensor position has been set correctly before calculating. The inclination can be aggregated to the output interval by taking the mean value of the samples in each interval.

## # Body Position (BodyPosition)

The DataAnalyzer uses the inclination obtained from the acceleration signals to classify the body position. Make sure the sensor location is set correctly. During one measurement the proband has to wear the sensor at the same sensor location. The following table shows which body positions can be detected for each sensor location:

Body position Coding hip chest thigh ankle wrist upper arm hip3 (*)
Unknown 0 X X X X
Lying supine 1 X X X X
Lying left 2 X X X X
Lying prone 3 X X X X
Lying right 4 X X X X
Upright 5 X X X
Sitting/lying 6 X
Standing 7 X X
Sitting 8 X

(*) Body position hip3 can be used to distiguish between sitting and standing at the sensor location hip, by accepting a lower classification accuracy. It is only available by using a custom configuration file.

For sensor placement see sensor location. Internally the body position is calculated in intervals of 4s. The output interval determines how frequently the body position is output by taking the mean value.

movisensLIVE: BodyPosition can be calculated live on the sensor.

## # Sedentariness

This algorithm is based on the activity metric Movement Acceleration Intensity and Body Position. It works for sensor locations thigh and hip3. An output interval is classified as sedentary if a sedentary body position (sitting/lying) is detected as well as movement acceleration intensity is below a ceratin threshold.

Activity class Coding
Unknown 0
sedentary 1
non-sedentary 2

## # Activity Class (ActivityClass)

This algorithm detects the activity class that’s occurring during each output interval using a combination of features calculated from the accelerometer data and the barometric air pressure. A White-Box decision tree was used to determine the activity class.

The following table shows the activity classes that can be detected:

Activity class Coding hip chest thigh ankle wrist upper arm hip3
Unknown 0 X X X X X X X
Lying 1 X X X X
Sitting/standing 2 X X X
Cycling 3 X X X X X X
Jogging 5 X X X X X X X
Walking 7 X X X X X X X
Sitting/lying 8 X
Standing 9 X
Sitting/lying/standing 10 X X
Sitting 11 X

For sensor locations "chest" and "hip" Activity class "unknown" is output if body position is "unknown".

Please check that the sensor location has been entered correctly. More on the different sensor locations can be found in the relevant sensor manual.

movisensLIVE: Activity Class can be calculated live on the sensor.

Literature and Validation:

## # Step Count (StepCount)

This parameter uses the acceleration signal to calculate the number of steps taken in a given output interval. At first the magnitude of the acceleration vector os calculated. After applying a low pass filter steps are detected by inspecting low and hing points while considering plausibility criteria. A minimum sequence of at least three consecutive steps is needed to allow step detection. The output interval determines how frequently StepsCount is output by summing all steps if each interval.

movisensLIVE: StepCount can be calculated live on the sensor.

## # Altitude (Altitude)

Using the signal from the barometric air pressure sensor, this parameter calculates the altitude above sea level. The altitude uses the barometric formula (exponential atmosphere) to reach its results, and thus, changes of air pressure due to changes in the weather can influence the output. The unit is meters [m].

The basis for measuring the change in altitude by means of a barometric pressure sensor is the fact that the atmosphere becomes denser at lower altitudes and thus the air pressure decreases with increasing altitude.

$H=\frac{T_{M,0}}{L_{M,0}} \cdot \left( \frac{P^{R \cdot L_{M,0}}}{P_0} - 1 \right)$
$P_0 = \text{Air pressure at Sea Level}$
$T_{M,0} = \text{Temperature at Sea Level}$
$L_{M,0} = \text{Temperature gradient}$

With this method, a relative height change can be measured with a resolution of up to 10 cm; statements about the absolute height (e.g. required for localization), are not possible due to the variations in air pressure due to weather conditions. The output interval determines how frequently altitude is output by taking the mean value.

## # Vertical Speed (VerticalSpeed)

This parameter calculates the mean of the vertical speed for each output interval. A positive value represents upward movement. The unit is [m/s] and the calculation uses the altitude calculated from barometric air pressure data.

## # PDF and Excel Reports

Several reports are available for the documentation of physical activity. PDF reports can be used as incentives for participants. The report ReportActivitySummaryExcel gives a high-level summary of the physical activity.

Report Type Description Aligned at full days
ReportActivityPdf PDF Detail report with plots of activity class, steps and movement intensity for each day (two pages per day). no
ReportActivityMediumPdf PDF Weekly plots for steps/day. Detailed plots of steps, activity class and movement intensity for each day (one page per day). no
ReportActivitySummaryPdf PDF Weekly plots of step, MET, weekly Table with activity classes no
ReportActivitySummaryShortPdf PDF Weekly plots of steps and activity class yes
ReportActivitySummaryExcel Excel One line per day for steps and activity classes yes
Last Updated: 4/22/2021, 8:37:48 AM