CONTEMPORARY APPROACHES TO AUTOMATED FALL DETECTION SYSTEMS
Keywords:
Arrhythmia detection, IoT based fall detection, data filtration, XAI, ML detection frameworks, systems based on DL detection, wearable sensors, SMOTE techniques, remote detection and prediction, GPS and GSM module, senior citizens.Abstract
It is observed that falls among senior citizens and patients with similar neurological disorders like epileptic seizures are very common these days. Therefore, falls are reported annually, which may lead to severe injuries. In certain cases, deaths are also reported, particularly when there is a lack of supervision or when the adults are residing alone in the UK or in European countries. To minimize these risks and enhance the independence of individuals, accurate fall detection systems have been proposed. These issues are addressed by presenting sensor based, IoT, ML, and DL based architecture by various authors. These systems employed classification techniques in order to classify fall and non fall events in real time. The data is cleaned by using bandpass filters and data preprocessing techniques. The authors used various classification models like SVM, KNN, 1DCNN, and 2DCNN along with the XAI techniques, which include LIME and SHAP. This paper generates a comprehensive review of the existing methodologies and highlights their pros and cons. Subsequent research of the existing studies of the past 5 years includes data security, exploration and development of systems that detect all seizure types, utilization of the advanced DL models, detection of falls in real time, and so much more. On the whole, this paper provides a comprehensive literature review that understands the existing systems, their challenges, and future directions, which helps to develop more reliable and efficient systems that would bring an evolution in the healthcare systems.
