Sensor-based human activity recognition: challenges ahead

Abstract

Human Activity Recognition (HAR) has explored a lot recently in the academia and industries for numerous applications. There are lots of progress in the domain of vision-based action or activity recognition due to the advent of deep learning models and due to the availability of very large datasets in the last several years. However, there are still a number of genuine challenges in vision-based HAR. On the other hand, sensor-based HAR has more constraints to decipher, and is still far from maturity due to various challenges. These core challenging issues are addressed in this chapter. The challenges regarding data collection issues have been discussed in detail. Prospective research works and challenges in the field of sensor-based activity recognition have been discussed in terms of new researchers perspective, possible developments in industries for experts, and smart IoT solutions in the medical sector.