Implementation of Artificial Intelligence Internet of Things (AIoT) Device for Activity Recognition

With everything being instant nowadays it results in someone without aware of experiencing a lack of movement. This can have a negative impact on health and can cause disease. To overcome this, currently there are many portable devices such as cell phones and smart watches that have functions as an activity monitor. However, it is still limited steps activity, even though human movement activities are not only manifested from steps. One of the problems is because the processing method still relies on processing simple digital signal processing (DSP). Other influencing factors is the processing component (central processing unit, CPU) that is used to have lightweight computing to pursue low device power consumption. This research develops a framework and prototype for custom human activity recognition (HAR) using inertial sensors. The microcontroller is used as a data processing unit so that it can recognize activities carried out. The method that will be used is a combination of digital signal processing and machine learning (ML), as well as equipped with an internet. As evaluation, hardware performance and activity recognition (3 basic table tennis movement) were tested. In the result, device prototype has maximum 245 mA draw current and up to 100% on accuracy of activity recognition. With IoT embed to the device its enable for further development and application usage.