Fall Prediction by a Spatio-Temporal Multi-Channel Causal Model from Wearable Sensors Data

Abstract

Predicting human falls from wearable devices is a complex task due to the inherent diversity and causality of multivariate physical changes, where each instance exhibits a unique style of motion events and their spatio-temporal causal dependencies. Consequently, we propose a multichannel causal model that utilizes the Granger causality test to explicitly delineate these internal configurations of motion events and their causal relationships from a spatio-temporal perspective. Particularly, our model incorporates a multi-head attention mechanism with a distillation component to capture the spatio-temporal dependencies among multiple channels of motion sensors in an end-to-end fashion. Empirical evaluations conducted on two benchmark datasets, as well as one in-house dataset collected by ourselves, indicate that our model significantly surpasses state-of-the-art approaches.