physical activity recognition using a wearbale accelerometer
DESCRIPTION
Public defense of my PhD thesisTRANSCRIPT
Physical activity and health
• A physically active lifestyle:
Reduces risk for adverse health conditions– Cardiovascular diseases, type 2 diabetes, cancer
Increases energy expenditure– Plays a role in the regulation of body weight
Physical activity and health
• A physically active lifestyle:
Reduces risk for adverse health conditions– Cardiovascular diseases, type 2 diabetes, cancer
Increases energy expenditure– Plays a role in the regulation of body weight
• How much physical activity is needed to gain health benefits?
Ene
rgy
expe
nditu
re
Measuring physical activity• Objective methods are required to establish the
relationship between:
Energy expenditureBody movement Health
Activity
Basal
Diet
Wearable accelerometer
• DirectLife activity monitor – unobtrusive, small – low battery consumption
Wearable accelerometer
• DirectLife activity monitor – unobtrusive, small– low battery consumption
• Measures body movement – activity counts Vertical
Forward
Sideward
Wearable accelerometer
• Predicts activity energy expenditure
Measured expenditure, MJ/day
Pre
dict
ed e
xpen
ditu
re,
MJ/
day
Explained variation = 53%
Physical activity recognition
Walking CyclingWalking
Standing
SittingFloor sweeping
Acceleration signal
Physical activity recognition
Cycling
Active Standing
Active Standing
Walking
Walking
Running Lying Sedentary
Decision tree model
Classification accuracy > 93%
Energy expenditure estimation
• Relation between body movement and energy expenditure depends on the activity type
Energy expenditure estimation
• Relation between body movement and energy expenditure depends on the activity type
• Activity recognition improves the estimation of activity energy expenditure– METD, metabolic cost of physical activity based on activity types
– activity counts
Behaviour and the activity level
• Objectively measured physical activity behaviour– Sample of the Dutch population
• Objectively measured physical activity behaviour– Sample of the Dutch population
Behaviour and the activity level
Walk 5.3%Cycle 1.6%Run 0.2%
Active standing
27%
Sedentary29%
Sleep37%
7%
• Objectively measured physical activity behaviour– Sample of the Dutch population
• On average, replacing 30 min/day of sitting with cycling increases daily energy expenditure by 10%
– Obese subjects measured before and after a diet
• After weight loss, activity energy expenditure decreases mainly because of the lower body weight carried during physical activity
• reducing 2 hours/day of sedentary time restores baseline activity energy expenditure
Behaviour and the activity level
Cardiovascular health
• In a group of obese subjects the time spent on active standing is associated with:– better index of heart rate variability
Increase in active standing, min/day
Adj
uste
d H
RV
, lo
g m
s2
Explained variation = 62%
Cardiovascular health
• In a group of obese subjects the time spent on active standing is associated with:– Lower plasma insulin concentration
Active standing, min/day
Pla
sma
insu
lin,
mm
ol/L
Explained variation = 18%
Conclusions
Physical activity recognition:
• was achieved using a single wearable accelerometer
• Improved the estimation of activity energy expenditure
• unraveled the relation between body movement and health with respect to cardiovascular diseases prevention