A smartphone-based fall detection system

doi:10.1016/j.pmcj.2012.08.003

Pervasive and Mobile Computing

Volume 8, Issue 6, December 2012, Pages 883-899

https://doi.org/10.1016/j.pmcj.2012.08.003 Get rights and content

Abstract

Falls are a major cause of injuries and hospital admissions among elderly people. Thus, the caregiving process and the quality of life of older adults can be improved by adopting systems for the automatic detection of falls. This paper presents a smartphone-based fall detection system that monitors the movements of patients, recognizes a fall, and automatically sends a request for help to the caregivers. To reduce the problem of false alarms, the system includes novel techniques for the recognition of those activities of daily living that could be erroneously mis-detected as falls (such as sitting on a sofa or lying on a bed). To limit the intrusiveness of the system, a small external sensing unit can also be used for the acquisition of movement data.

Introduction

A fall event is one of the main factors that influence the physical and psychological health of an elderly person. Injuries related to falls include physical damages like skin abrasions, bone fractures and general connective tissue lesions [1], [2]. A fall also has dramatic psychological consequences, since it drastically reduces the self-confidence and independence of affected people. This may contribute to future falls with more serious outcomes or it may lead to a decline in health. Other frequent consequences include early nursing home admission and continuous fear of falling, lowering the quality of everyday life [3]. The consequences of a fall event depend also on the long-lie period, i.e. the time interval during which the person remains involuntarily on the ground after the fall [4]. Therefore, it is of fundamental importance to provide quick support to the injured people as soon as a fall happened.

The simplest solution to the fall detection problem consists of providing people with a Personal Emergency Response System (PERS), a small, light-weight and battery-powered device with a “help” button that can be carried on a belt, in a pocket, on a necklace or on a wrist band. This kind of device also embeds a radio transmitter which is able to connect to the user’s home telephone and to dial preselected numbers in case of emergency. Many of these systems have been successfully deployed in several countries and require almost no configuration [5]. However, they suffer from a major issue: the need for the user to press a button. Unfortunately, it is common that after a fall a person is unable to perform even this simple action, for example because of a loss of consciousness or because the final lying position could prevent the victim from being able to reach the button.

Thus, in the last few years, research about systems for the automatic detection of falls gained momentum, pushed by the growing number of elderly citizens in a large fraction of the world [6], [7], [8], [9], [10], [11]. The techniques for the automatic detection of falls can be substantially divided into two categories [12]. The first category includes the approaches based on instrumenting the environment; examples include equipping rooms with cameras able to track the movements of people or placing pressure sensors in specific areas (e.g., in the vicinity of beds). The second category includes techniques based on wearable sensors: accelerometers and/or gyroscopes are used to collect kinematic information about the monitored person and then to detect falls. The advantage of using wearable sensors is that almost no installation or set-up is required and the system is immediately available for deployment. Moreover, the area of operation is not limited to the instrumented spaces as the system can be carried by users wherever they go. This enlarges the number of users and situations, ideally including all the activities that expose people to long periods of being alone with a high risk of falling. Nevertheless, the user is required to wear at least a device and this can pose some intrusiveness and usability concerns. In other situations, automatic fall detection can be integrated within the monitoring system dedicated to specific pathologies [13].

Many automatic fall detection systems suffer from the problem of false alarms, caused by some fall-like activities of daily living (ADLs), such as sitting on a sofa or lying on a bed. For this reason, in our approach to fall detection we devoted special attention to the study of the acceleration signal produced by fall-like ADLs and to the design of novel filtering techniques [14]. In this paper, we describe the design rationale and the implementation of a fall detection system based on wearable sensors. The system relies on commercially available smartphones and is capable of automatically sending an alarm message to the caregivers in the case of a fall. The acquisition of kinematic data can be carried out either using the accelerometer available on many smartphones or using an external sensing unit. The usability of the system has been confirmed by a set of interviews with some aged people, while its performance, in terms of precision and recall, has been evaluated both in lab sessions and through continuous monitoring of three subjects (including indoor and outdoor activities). A comparison with similar existing fall detection techniques is also reported.

Section snippets

Requirements and architectural guidelines

A fall detection system can be useful for people working or doing recreational activities in isolated places with a high risk of falls, such as country workers, mushroom hunters, or skiers. However, the category of people that can benefit more significantly from a fall detection system is the elderly. In fact, in the last few years life expectancy has increased making a larger fraction of the population more prone to falls. Unfortunately, the injuries due to falls are a major cause of

Implementation

This section describes the hardware platform used to build a prototype of the system and the software running on the smartphone and on the external sensing unit when provisioned.

Detection of real falls and filtering of fall-like ADLs

The building blocks of the Classification Engine are shown in Fig. 4. As said in the previous section, fall-like events that are not discarded through the Activity Test are forwarded to the Classification Engine. The input to the engine is a vector of acceleration magnitude values collected in the interval $[- 2200, 1000] ms$ , centered at the instant of the peak exceeding the $3 g$ threshold. These values are fed into a Feature Extractor whose task is to reduce the number of input values. This is

Performance evaluation

The observed behavior of a fall detection system is represented by four possible situations: true positive (TP), a fall occurred and the system correctly detects it; false positive (FP), the system declares a fall event, but it was instead a normal ADL; true negative (TN), the system correctly classifies a fall-like event as an ADL; false negative (FN), a fall occurred, but the system does not detect it. System reliability can then be evaluated through the following indexes [27]: $s e n s i t i v i t y = T P /$

Comparison with relevant existing techniques

This section reports the results of the quantitative comparison of our fall detection approach with some relevant techniques described in recent literature. In particular, we selected a set of methods that have been conceived to operate using a wearable waist-mounted accelerometer. All of these methods have been re-implemented following the description available in the published papers. Then, they have been fed with our traces of intentional falls and fall-like activities, in order to evaluate

Smartphone-based fall detection systems

Several smartphone-based fall detectors have been developed so far. This confirms that the idea of using a smartphone as the basis for a fall detection system is sound. All the systems make use of the embedded accelerometer to detect a fall, but use different algorithms. The systems span from a simple threshold-based fall detector to a wavelet analyzer.

The authors of PerfallD [34] propose a comprehensive system to integrate fall detection and emergency communication. They designed two

Conclusion

A fall is an ill-defined process, therefore it is not an easy task to describe it thoroughly. This implies that identifying signal patterns, and then defining parameters and filters that can be applied to distinguish falls from common activities, requires significant effort. Moreover, any technique based on analytical-only methods would require further refinement to be tailored to new users. For these reasons, a machine learning approach can be successfully applied, together with analytical

Acknowledgments

The work described in this paper has been partly supported by the Fondazione Cassa di Risparmio di Lucca 2010 project “Fido: a fall detection alarm system for elderly people” and by the MIUR-PRIN 2008 project “Cloud@Home: a New Enhanced Computing Paradigm”.

References (39)

A.K. Bourke et al.
Evaluation of a threshold-based tri-axial accelerometer fall detection algorithm
Gait & posture
(2007)
A.K. Bourke et al.
A threshold-based fall-detection algorithm using a bi-axial gyroscope sensor
Medical engineering & physics
(2008)
M. Avvenuti et al.
An application adaptation layer for wireless sensor networks
Pervasive and Mobile Computing
(2007)
M. Kangas et al.
Comparison of real-life accidental falls in older people with experimental falls in middle-aged test subjects
Gait & Posture
(2012)
N. Noury et al.
A proposal for the classification and evaluation of fall detectors
IRBM
(2008)
A. Bourke et al.
Evaluation of a threshold-based tri-axial accelerometer fall detection algorithm
Gait and Posture
(2007)
M. Kangas et al.
Comparison of low-complexity fall detection algorithms for body attached accelerometers
Gait & Posture
(2008)
Public Health Agency of Canada, Report on Seniors’ falls in Canada,...
S. Sadigh et al.
Falls and fall-related injuries among the elderly: a survey of residential-care facilities in a Swedish municipality
Journal of community health
(2004)
R. Cumming et al.
Prospective study of the impact of fear of falling on activities of daily living, SF-36 scores, and nursing home admission
The Journals of Gerontology Series A: Biological Sciences and Medical Sciences
(2000)

D. Wild et al.

How dangerous are falls in old people at home?

British Medical Journal

(1981)

Philips, Lifeline - the trusted medical alert service provider, http://www.lifelinesys.com,...

S. Abbate, M. Avvenuti, P. Corsini, J. Light, A. Vecchio, Monitoring of Human Movements for Fall Detection and...

N. Noury, A. Fleury, P. Rumeau, A. Bourke, G. Laighin, V. Rialle, J. Lundy, Fall detection - principles and methods,...

M. Kangas, A. Konttila, I. Winblad, T. Jamsa, Determination of simple thresholds for accelerometry-based parameters for...

Q. Li, J.A. Stankovic, M.A. Hanson, A.T. Barth, J. Lach, G. Zhou, Accurate, fast fall detection using gyroscopes and...

X. Yu, Approaches and principles of fall detection for elderly and patient, in: Proceedings of the 10th International...

M. Avvenuti, C. Baker, J. Light, D. Tulpan, A. Vecchio, Non-intrusive patient monitoring of alzheimer’s disease...

S. Abbate, M. Avvenuti, G. Cola, P. Corsini, J.V. Light, A. Vecchio, Recognition of false alarms in fall detection...

Cited by (295)

Sensors and algorithms for locomotion intention detection of lower limb exoskeletons
2023, Medical Engineering and Physics
In recent years, lower limb exoskeletons (LLEs) have received much attention due to the potential to help people with paraplegia regain the ability of upright-legged locomotion. However, one major hindrance to converting prototypes into actual products is the lack of a balance recovery function. Locomotion intentions can be the first step for balance assistance. Therefore, its significance continues to grow. Many researchers focus on this topic, but there is a lack of a general discussion on the research phenomenon. Therefore, the purpose of this work is to systematize these data and benefit future research. This review is divided into two parts, the location of sensors/devices and the evaluation criteria of algorithms, which are the main components of locomotion intentions. We found that sensor/device placement is still concentrated in the lower limbs, but most researchers have found the importance of the chest. The peak power of the signal collected from the chest may be overestimated because it undergoes higher vertical velocity and acceleration during a rotation. However, despite the differences in peak power between the upper and lower back, high correlations were found for the tasks, especially from sitting to standing. Since peak power is based on vertical acceleration and velocity, it can be considered a metric that is more robust to changes in sensor location. Therefore, data acquisition from the chest is effective. In this paper, it is pointed out that sensors placed on the chest may have a tendency to change, as some researchers have realized in the field of locomotion intention recognition. In the evaluation criteria, we also found that deep learning algorithm (such as Back Propagation Artificial Neural Network (BPANN)) is outstanding, and Support Vector Machine (SVM) is the most cost-effective algorithm. In terms of accuracy, sensitivity, and specificity, BPANN achieved nearly 100%. SVM has different types; the best one achieves 98% accuracy, 100% sensitivity, and 100% specificity. But it also has 87.8% accuracy, which is not stable. Convolutional Neural Networks (CNN) can be used for image classification and have an accuracy of around 87%. Compared to the above two algorithms, CNN may have lower performance. Other algorithms also have higher accuracy, sensitivity, and specificity. These evaluation criteria, however, were not all ideal at the same time. Based on these results, we also point out the existing problems. In general, the application of these algorithms to LLE can contribute to its intention recognition, which can be helpful in balancing research. Finally, this can help make LLE more suitable for daily use.
An analytical comparison of datasets of Real-World and simulated falls intended for the evaluation of wearable fall alerting systems
2022, Measurement: Journal of the International Measurement Confederation
Automatic fall detection is one of the most promising applications of wearables in the field of mobile health. The characterization of the effectiveness of wearable fall detectors is hampered by the inherent difficulty of testing these devices with real-world falls. In fact, practically all the proposals in the literature assess the detection algorithms with ‘scripted’ falls that are simulated in a controlled laboratory environment by a group of volunteers (normally young and healthy participants). Aiming at appraising the adequacy of this method, this work systematically compares the statistical characteristics of the acceleration signals from two databases with real falls and those computed from the simulated falls provided by 18 well-known repositories commonly employed by the related works. The results show noteworthy differences between the dynamics of emulated and real-life falls, which undermines the testing procedures followed to date and forces to rethink the strategies for evaluating wearable fall detectors.
A comprehensive comparison of accuracy and practicality of different types of algorithms for pre-impact fall detection using both young and old adults
2022, Measurement: Journal of the International Measurement Confederation
Citation Excerpt :
It has been reported that long lies accounted for 30 % of fall accidents and people experienced a long lie were more likely to suffer subsequent serious injuries [6]. The deployment of fall detection systems can either prevent fall-induced injuries by activating protective devices such as inflatable airbags [7] or avoid long lies by notifying the occurrence of falls through communication devices such as smartphones [8]. In both types of systems, fall detection algorithm is the core to make them effective.
This study aims to comprehensively compare the accuracy and practicality of three different types of algorithms for pre-impact fall detection using both young and old subjects. Threshold-based, conventional machine learning (SVM) and deep learning (ConvLSTM) algorithms were compared. Results showed that ConvLSTM had an accuracy of 99.16 % (sensitivity: 99.32 %, specificity: 99.01 %) and an averaged lead time of 403 ms on young subjects, which outperformed SVM (97.16 %, 385 ms) and much superior to the threshold-based algorithm (89.06 %, 333 ms). In addition, latency tests on an embedded device showed that the Lite model of ConvLSTM had a low latency of 2.1 ms, which was comparable to the threshold-based algorithm (<1 ms) but much lower than SVM (86.9 ms). The feasibility and effectiveness of applying algorithms trained on young subjects to old subjects were also validated. These findings suggested that ConvLSTM has great potential for detecting pre-impact falls and preventing fall-related injuries.
Towards effective detection of elderly falls with CNN-LSTM neural networks
2022, Neurocomputing
Fall detection is a very challenging task that has a clear impact in the autonomous living of the elderly individuals: suffering a fall with no support increases the fears of the elderly population to continue living by themselves. This study proposes the use of a non-invasive tri-axial accelerometer device placed on a wrist to measure the movements of the participant. The novelty of this study is two fold: on the one hand, the use of a Long-Short Term Memory Neural Network (LSTM) for classification of the Time Series and, on the other hand, the proposal of a novel data augmentation stage that introduces variability in the training by merging the Time Series gathered from both human activities of daily living. The experimentation shows that the combination of a LSTM model together with the data augmentation produces more robust and accurate models that perfectly cope with the validation stage; the high impact fall event detection can be considered solved.
Edge Computing Transformers for Fall Detection in Older Adults
2024, International Journal of Neural Systems
A Physiological Sensor-Based Android Application Synchronized with a Driving Simulator for Driver Monitoring
2024, arXiv

View all citing articles on Scopus

View full text

A smartphone-based fall detection system

Abstract

Introduction

Section snippets

Requirements and architectural guidelines

Implementation

Detection of real falls and filtering of fall-like ADLs

Performance evaluation

Comparison with relevant existing techniques

Smartphone-based fall detection systems

Conclusion

Acknowledgments

Gait & posture

Medical engineering & physics

Pervasive and Mobile Computing

Gait & Posture

IRBM

Gait and Posture

Gait & Posture

Falls and fall-related injuries among the elderly: a survey of residential-care facilities in a Swedish municipality

Journal of community health

Prospective study of the impact of fear of falling on activities of daily living, SF-36 scores, and nursing home admission

The Journals of Gerontology Series A: Biological Sciences and Medical Sciences

How dangerous are falls in old people at home?

British Medical Journal