Design and Analysis of Fall Detection Method Using Machine Learning in Healthcare

Abstract

Fall is a major threat to the health and life of the elders. A Fall Detection System newline(FDS) assists the elders by identifying falls accurately and save their lives. Machine Learning (ML) based FDS has turned into a major research area due to its capability to assist the elders automatically. The efficiency of a fall detection system depends on its strength to identify the newlinefall from non-fall accurately. The initial fall detection scheme depends on the threshold-based classification to classify the fall from the Activity of Daily Living (ADL), but this classification method has failed to reduce the false alarm rate, which raises a question on the efficiency of the technique. Fall Detection (FD) system tends to monitor the fall events with restricted movement patterns and provides alerts to detect actions corresponding to human falls. Based on high-level features, the resultant information often requires well detected results like activity monitoring, detection, and classification. We have also focused on fall detection model through convolutional Long Short-Term Memory (ConvLSTM) algorithm with varying newlineoptimizers performed on video dataset to evaluate model s performance. In this thesis we have developed and proposed a novel approach for efficient fall detection. The proposed ensemble based FDS focuses to improve the fall detection accuracy by choosing an optimal classifier newlinewith greedy algorithm based majority voting approach. In this ensemble model four machine newlinelearning algorithms namely Support Vector Machine (SVM), K-Nearest Neighbour (KNN), newlineDecision Tree and Deep LSTM are used for classification. The greedy algorithm based newlinemajority voting employs a search based approach namely forward search, backward search and recovery search, with the objective to select optimal classifiers. The search pattern is mapped to selection criteria while choosing the classifiers. The forward search adds the optimal classifiers, while the backward search removes the classifier based on the selection criteria.