Overview / Usage

In recent years, there have been a surge in ubiquitous technologies such as smartwatches and fitness trackers that can track human physical activities effortlessly. These devices have enabled common citizens to track their physical fitness and encourage them to lead a healthy lifestyle. Among various exercises, walking and running are the most common activities people do in everyday life, either to commute, exercise, or do household chores. While performing these activities, the speed at which a person walks or runs is an essential factor to determine the intensity of activity. Therefore, it is important to measure walking/running speed to estimate the burned calories along with preventing the risk of soreness, injury, and burnout. Existing wearable technologies use a GPS sensor to measure speed, which is highly energy inefficient and does not work well indoors. To solve this problem, we design, implement, and evaluate a convolutional neural network-based algorithm that leverages data from accelerometer and gyroscope sensors in a wrist-worn device to detect speed with high precision. We have also evaluated various other machine learning algorithms to compare our results.

Methodology / Approach

This study aims to build a precise algorithm to detect the speed of walking and running related activities. First, it collects accelerometer and gyroscope sensory data from wrist-worn device. Then it tests different feature engineering techniques, machine learning and neural network architectures. Overall the study shows the limitations associated with different algorithms and the trade-off between model complexity and model accuracy.

Technologies Used

Python, Sci-kit learn, C#, Android Application, Tensorflow

Other links

• EdgeDL conference paper
• Master Thesis