Implementation of a SVM on an Embedded System: A Case Study on Fall Detection

Edge Computing seeks to bring Machine Learning as close as possible to the source events of interest, providing an almost instant interpretation to data acquired by sensors giving sense to raw data while addressing concerns of particular applications such as latency, privacy and server stress relieve. Due to a lack of research on this particular type of application, we are faced with difficulties both in software and hardware as embedded systems are known to possess serious limitations on its available processing resources. To address this, we make use of the concepts of edge computing and offline programming to accomplish a reliable machine learning model deployment on the microprocessor. By studying real case problem, we can get measurements on the resources required by such an application as well as its performance. In this study, we address the implementation of such an application in an embedded system focusing on the detection of human falls. © 2020, Springer Nature Switzerland AG.