Towards Putting Deep Learning on the Wrist for Accurate Human Activity Recognition

The use of smart wearables enables regular monitoring of human activities, which can lower risks of health complications due to cardiovascular diseases, diabetes, etc. The signal data generated by the accelerometer and gyroscope of the inertial measurement unit (IMU) within wearables, aid in recognizing motion. In the past, researchers have used Signal Processing techniques for human activity recognition (HAR). However, these techniques are not adaptive to variations of the same activities in different people. Machine Learning and Deep Learning are comparatively adaptive to variations but require more computational power, especially Deep Learning. To circumvent high computational requirements, state-of-the-art solutions typically deploy the learning algorithms on a platform with more compute power such as a smartphone, or even Cloud. In this paper, we propose a classification model for HAR that outperforms state-of-the-art prediction accuracy on raw signal data and is suitable for deployment in resource-constrained environments of smart wearables and IoT devices.