Paper Title
Multi-Headed Attention Network Based Fall Detection System Using a Single IMU

Abstract
Fall detection is an imperative task within the area of human activity recognition and cyber physical systems. Given that falls are one of the leading cause of injuries for elderly people, the development of reliable fall detection systems has garnered keen interest by the research community. This paper presents a multi-headed attention network that uses CNNs for fall detection purposes. To perform this task, experiments have been carried out windowed segments of two different publicaly available datasets, SisFall and KFall. The results indicate that the network performs suitably well in both cases, achieving an average F1 score of more than 98% for KFall and more than 97% for SisFall. This work will serve as a basis for real-time deployable fall detection systems. Keywords - Fall Detection, Internet of Things (IoT), Transformers,CNN