CapsFall: Fall Detection Using Ultra-Wideband Radar and Capsule Network

Abstract

Radar technology for at home health-care has many advantages such as safety, reliability, privacy-preserving, and contact-less sensing nature. Detecting falls using radar has recently gained attention in smart health care. In this paper, CapsFall, a new method for fall detection using an ultra-wideband radar that leverages the recent deep learning advances is proposed. To this end, a radar time series is derived from the radar back-scattered matrix and its time-frequency representation is obtained and used as input to the capsule network for automatic feature learning. In contrast to other existing methods, the proposed CapsFall method relies on multi-level feature learning from radar time-frequency representations. In particular, the proposed method utilizes a capsule network for automating feature learning and enhancing model discriminability. The experiments are conducted using a set of radar signals collected from ten subjects when performing various activities in a room environment. The performance of the proposed CapsFall method is evaluated in terms of classification metrics and compared with those of the other existing methods based on convolutional neural network, multi-layer perceptron, decision tree, and support vector machine. The results show that the proposed CapsFall method outperforms the other methods in terms of accuracy, precision, sensitivity, and specificity values. INDEX TERMSBiomedical signal processing, smart home care, fall detection, classification. MIODRAG BOLIC (M'04-SM'08) received the Ph.D. degree in electrical engineering from Stony Brook University, USA, in 2004. Since 2004, he has been with the University of Ottawa, Ottawa, Canada, where he is currently a Professor with the School of Electrical Engineering and Computer Science. He published about 70 journal papers, four book chapters, and edited one book. His current projects are related to development of contactless monitoring systems using radars and cameras. His current research interests include machine learning, and biomedical signal processing and instrumentation. SREERAMAN RAJAN received the B.E. degree from Bharathiyar University, India, the M.Sc. degree from Tulane University, USA, and the Ph.D. degree from the University of New Brunswick, Canada. He was a Senior Defence Scientist with the Defence Research and Development Canada and an Adjunct Professor with the School of Electrical Engineering and Computer Science, University of Ottawa. He has worked in nuclear science and engineering research, fiber optical communication modules and systems, and non-invasive medical devices industries. He is currently a Tier 2 Canada Research Chair with the

Unobtrusive human activity classification based on combined time‐range and time‐frequency domain signatures using ultrawideband radar

“…[ with fivefold cross-validation (AC = 91.1%), compared with the results obtained in CapsFall proposed approach in Sadreazami et al [38], in which only the TF domain (spectrogram) was used as input for a capsule module network (automatic feature learning) to differentiate fall and non-fall activities. The comparison is done only between the Caps-Fall and the first-stage classification of the proposed method to compare performance regarding the classification of two classes.…”

“…In work by the same authors [37], the features were learned automatically using DCNN from the radar time series data. The authors [38] also proposed a capsule network-based fall detection (CapsFall) method in which the TR representation of the radar time series was fed into a CapsFall method for automatic feature learning. Du et al [39] introduced a three-dimensional (3D) DL framework for human motion analysis in which reflected radar echoes were transformed into range-Doppler points to obtain the discrete representation of motion trajectory.…”

High Accuracy WiFi-Based Human Activity Classification System with Time-Frequency Diagram CNN Method for Different Places

“…Literature [ 37 , 38 ] have discussed the usage of radar technology for detecting falls. Study [ 37 ], proposes CapsFall, based on Ultra-Wideband (UWB) radar which relies on multi-level feature learning from radar time-frequency representations.…”

“…The proposed solution outperforms the other methods. Paper [ 38 ] propose a learning model that combines convolutional layers and convolutional long short term memory (ConvLSTM) to extract robust spatiotemporal features for fall detection by using monostatic radar. In a word, UWB requires some specific hardware devices to detect falls and is not common in general households compared with WiFi devices.…”

Bi-LSTM Network for Multimodal Continuous Human Activity Recognition and Fall Detection

“…Neural network-based classification methods, in particular, CNN (Convolutional Neural Network) and autoencoders have attracted a lot of attention and showed to generally outperform conventional classifiers in terms of classification accuracy, at the price of additional training complexity. H. Sadreazami et al [23] proposed a CNN-based Capsule network to identify the fall accidents through ultra-wide band radar and the results indicated that it over performed SVM with different kernel [24] also pointed that SVM plus 2-D PCA (Principle Component Analysis) and CNN (RadarNet) showed better performance than other classifiers in activity classification, subject recognition and outdoor localization. S. Gurbuz et al [25] compared the walking pattern recognition of three different radar sensors and one sonar using a broad range of features.…”