A Novel On-Wrist Fall Detection System Using Supervised Dictionary Learning Technique

Abstract

Wrist-based fall detection system provides a very comfortable and multi-modal healthcare solution, especially for elderly risking falls. However, the wrist location presents a very challenging and unstable spot to distinguish falls among other daily activities. In this paper, we propose a Supervised Dictionary Learning approach for wrist-based fall detection. Three Dictionary learning algorithms for classification are invoked in this study, namely SRC, FDDL, and LRSDL. To extract the best descriptive representation of the signal data we followed different preprocessing scenarios based on accelerometer, gyroscope, and magnetometer. A considerable overall performance was obtained by the SRC algorithms reaching respectively 99.8%, 100%, and 96.6% of accuracy, sensitivity, and specificity using raw data provided by a triaxial accelerometer, accordingly overthrowing previously proposed methods for wrist placement.

Energy-Aware IoT-Based Method for a Hybrid On-Wrist Fall Detection System Using a Supervised Dictionary Learning Technique

“…This paper is an extension of previous works [ 16 , 17 ]. In those works, we first introduced a Supervised Dictionary Learning approach for wrist-based fall detection [ 16 ].…”

“…This paper is an extension of previous works [ 16 , 17 ]. In those works, we first introduced a Supervised Dictionary Learning approach for wrist-based fall detection [ 16 ]. The Sparse Representation-based Classifier (SRC) algorithms obtained an impressive overall performance reaching 99.8% accuracy based on an optimized solution using only a triaxial accelerometer.…”

“…PRDB: Taking into account the obtained result in our previous work [ 16 , 17 ], the Sparse Representation-based classifier has proved to have a better offline performance and robustness facing noises when trained on multiple situations of the Quadros et al [ 11 ] dataset. Thus, we will be conducting our work to compare the best SDL performance data with our proposed enhanced Algorithm 1 in real-life simulation using a raw accelerometer.…”

“…We have limited the FIFO register buffer to 200 samples for each axis ( buff_thresh ), with a 100 Hz sampling rate; SWB: We previously fixed a window size of 4 s, knowing that a fall can take two to three seconds. We assume that, in order to double predict the occurrence of a fall, we have set a FOSW to 75%; PRDB: Taking into account the obtained result in our previous work [ 16 , 17 ], the Sparse Representation-based classifier has proved to have a better offline performance and robustness facing noises when trained on multiple situations of the Quadros et al [ 11 ] dataset. Thus, we will be conducting our work to compare the best SDL performance data with our proposed enhanced Algorithm 1 in real-life simulation using a raw accelerometer.…”

“…Based on our previous works [ 16 , 17 ], we assume that the top three best SDL offline performances are achieved by a Sparse Representation-based Classifier (SRC) [ 33 , 34 ], Low-Rank Shared Dictionary (LRSDL) [ 35 , 36 ], and Dictionary Learning with Structured Incoherence (FDDL) [ 37 ]. This is obtained when the methods are trained on 75% of the dataset and tested on 25%.…”