Energy Estimation of Treadmill Walking Using On-Body Accelerometers and Gyroscopes

Background -Regular physical activity (PA) has many health benefits -Walking: most common activity among the physically active -Energy expenditure due to walking assists assessments of PA levels and determine effectiveness of interventions -Limitations of current measurement of PA by accelerometry -Imprecision–Due to proprietary methods to convert linear accelerations in to epoch-based counts -Incompleteness-Accelerometers do not completely describe human body movement

Purpose -Develop a map from movement to energy expenditure (VO₂ consumption) for treadmill walking -Capture movement with streaming inertial sensor data -Sensor contains tri-axial accelerometer(MMA7260Q) and gyroscope (with 2 Invensense ID G300 gyroscopes) -Develop a probabilistic mapping between inertial data and energy expenditure -Compare accuracy with Actigraph GT1M counts, GT1M is a commercially available tri-axial accelerometer that reports acceleration using a proprietary unit-“counts” -Explore utility of tri-axial accelerometers and gyroscopes in estimating energy expenditure due to walking

Results Figure 2: Illustration of correlation between IMU generated feature vector and GT1M readings for the X-axis acceleration signal for walking at five different speeds (color coded by VO₂ consumption). IMU feature vectors showed a strong linear correlation with the GT1M generated counts (.9787 ± .0089 for the X-axis and .9141 ± .0460 for the Y-axis) which means that features are related to each other by a simple linear transformation.

Figure 3: Illustration of comparison of IMU feature vector with Actigraph counts averaged across all participants when trained using BLR. Actigraph counts averaged across all participants. Each bar shows average RMS error±1 standard deviation. RMS error using the IMU feature vector is smaller than that from using Actigraph counts. This is reasoned to be because the sensor is more closely attached to the body than the GT1M leading to better motion capture. The IMU also has a higher sampling rate and accounts for zero bias.

Figure 4: Illustration of effect of introducing gyroscopic and z-axis in formation to train data using BLR. Each bar shows average RMS error ±1 standard deviation. While individually, training using accelerometer (all 3 axes) or gyroscopic data (all 3 axes) shows a higher error, combining data from multiple streams lowers the error. This is attributed to redundancy in movement in formation being captured through gyroscopes.