Multi-sensor Exercise-based Interactive Games for Fall Prevention and Rehabilitation

A. Santos, V. Guimarães, N. Matos, J. Cevada, C. Ferreira, I. Sousa

9th International Conference on Pervasive Computing Technologies for Healthcare (Pervasive Health), 2015

According to statistics, one in every three adults ageing 65 or older falls every year. Every fall may lead to long-term consequences due to fractures or even neurological damages. These consequences have severe impact in their quality of life, independence and confidence, ultimately increasing the risk of early death. Moreover, the risk of falling increases as age advances. Fortunately, several studies reveal that specific exercise programmes may help in reducing the risk of falling if performed correctly and frequently. However, user engagement and adherence to these programmes are still low mainly due to motivational factors, since interventions are usually long, unadapted and unchallenging. In this paper, a new solution is presented, which uses the concept of interactive games using motion sensors to tackle low adherence (through gaming motivation) and help in physical rehabilitation and reduce fall risk on elderly people by improving balance, muscle strength and mobility. It is intended to be used in community or domestic unsupervised contexts and supports relatively inexpensive sensing equipment (currently Kinect, Leap Motion, Orbotix Sphero and Smartphones) and common platforms (desktop and mobile). Tests were already undertaken with several individuals ageing 65 or more and the results were analysed and discussed, being generally positive, despite some issues in the movement detection algorithms.

Fall Detection Algorithm Using the Accelerometer of the Smartphone

J. Silva, B. Aguiar, T. Rocha, F. Sousa, and I. Sousa

Oral presentation at the International Conference on Ambulatory Monitoring of Physical Activity and Movement (ICAMPAM), Limerick, Ireland, 2015

Introduction: Falls are a frequent occurrence in the elderly population. An automatic fall detection has the potential to improve the assistance response time after a fall event so that appropriate assistance can be provided right after the fall.

Methods: This paper presents an algorithm for fall detection based on the smartphone built-in inertial sensors. The algorithm was implemented as a state machine with five stages (Stable, Unstable, Falling, Impact and Unconscious Watcher) as shown in the attached Figure. A validation dataset was collected with 28 young subjects, obtaining 3024 samples of falls (N=1369) and non falls (N=1655) collected according to the protocol defined by Noury et al. Moreover, a field trial was conducted in which a set of 32 users divided in two groups, one of 28 subjects (mean age of 60 years old) and the other of 4 subjects (with more than 65 years old), were continuously carrying the smartphone with enabled fall detection in the pocket for 15 days and 3 months, respectively. When the user’s phone is in his pocket, the accelerometer of the smartphone is continuously screened for fall events.

Results: For the validation dataset, the sensibility of the fall detection algorithm is 97.58%. In the field trial, for a continuous usage of the smartphone, we obtained 15 false positives (FP) for the first group, yielding 1 FP per day. For the elderly group an average of 0.45 FP were obtained per day.

Discussion and Conclusion: The proposed solution is a reliable system for fall detection that can help older users in their daily living with a false alerting of 0.45 non falls events per day. Further improvements should be considered to decrease the number of false alarms per day without causing an impact on the accuracy of the system and the capability to detect real fall events, especially for younger more active users, that are likely to have false alarms.

Energy Expenditure Estimation using the Accelerometer of the Smartphone

J. Silva, S. Carneiro, B. Aguiar, T. Rocha, and I. Sousa

Poster presentation at the International Conference on Ambulatory Monitoring of Physical Activity and Movement (ICAMPAM), Limerick, Ireland, 2015

Introduction: Accurate and objective assessment of physical activities and energy expenditure in real life conditions is an important metric for monitoring the health condition. There are several methods to estimate the energy expenditure using accelerometers, including the counts per minute estimation and speed-based methods. We propose an alternative method to estimate metabolic equivalents (METs) based on a linear regression derived from the values of METs retrieved from the oximeter and the feature root mean square (fRMS) of the magnitude of the accelerometer signal.

Methods: The inertial sensors data was collected from a group of 13 young and healthy subjects, aging 33 ± 9 years, in a laboratory environment. The subjects performed an incremental speed treadmill protocol while carrying one smartphone on each side of the belt and an oronasal oximeter mask. Signal processing techniques, feature extraction and selection methods were applied to the accelerometer signal, which was recorded with a sampling frequency of 33.33 Hz. The fRMS was found to be highly correlated with the METs value. A new model based on this feature was developed and the results were compared against indirect calorimetry outputs .

Results: Applying the Wilcoxon Signed Rank Test, statistically significant differences (p > 0.05) were not found between the indirect calorimetry and the proposed model. This regression has a normalized root mean squared error (NRMSE) of 20% and a Pearson correlation coefficient of 0.90 for the METs estimation, as shown in the Figure attached.

Discussion and Conclusion: The model derived was found to be a suitable method to estimate the energy expenditure using a smartphone. Moreover, the other methods found in the literature are dependent on the performance of the algorithms to calculate the counts per minute or to estimate the speed, which may lead to additional cumulative errors.

Accelerometer-based methods for energy expenditure using the smartphone

S. Carneiro, J. Silva, B. Aguiar, T. Rocha, I. Sousa, T. Montanha, J. Ribeiro

IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2015, pp. 151-156

Quantifying the energy expended during physical activity is an important metric to evaluate the quality and progress of individual training. There are several methods to estimate the energy expenditure using accelerometers, the most common are based on calculating counts per minute from the accelerometer signal to determinate the activity intensity in terms of metabolic equivalents (METs). This paper compares three methods to estimate the energy expenditure, the first has been proposed in a previous study and the last two are based on linear regressions derived from the data collected, one using speed, and the other using the feature root mean square (fRMS) of the magnitude of the accelerometer signal. These models were compared with indirect calorimetry outputs of energy expenditure during an incremental speed treadmill protocol. No statistically significant differences (p>0.05) were found between the indirect calorimetry and the model derived using the RMS feature, obtaining a normalized error of 20% for the METs estimation. In conclusion, this was found to be the most suitable method to estimate the energy expenditure from accelerometer data collected using a smartphone placed in the belt.

A Smartphone-Based Fall Risk Assessment Tool: Testing Ankle Flexibility, Gait and Voluntary Stepping

V. Guimarães, D. Ribeiro, L. Rosado, and I. Sousa

IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2014, pp. 1–6

Falls are a frequent occurrence in elderly population, contributing significantly to injuries and decreased quality of life. However, through evidence-based interventions for fall prevention, specific risk factors can be modified, offering an opportunity to reduce the number of falls among older persons. An enhanced knowledge of the nature of the risks is of paramount importance to enable a proper design of preventive schemes. In this paper, indicators of declines in balance and mobility are evaluated using a smartphone-based system in alternative to traditional methods such as force platforms and cameras. Fall risk assessment tests were adapted to the smartphone enabling it to test Ankle Flexibility, Gait and Voluntary Stepping. Experimental results show a good correlation of most of the variables derived from the tests using the smartphone and force platforms or cameras. In conclusion, the implementation of these tests in a smartphone platform to measure fall risk-related features supports the feasibility of a valuable alternative to the traditional technology used in dedicated laboratories for the analysis of movement.

Smartphone Based Fall Prevention Exercises

B. N. Ferreira, V. Guimarães, and H. S. Ferreira

IEEE 15th International Conference on e-Health Networking, Applications Services (Healthcom), 2013, pp. 643–647

Falling is a very serious problem for our society, as it affects one out of three older adults. Currently, this is a well-known problem and therefore multiple ICT-based solutions for falls management exist. In addition, a small part of them are said to help preventing falls, but most of the reviewed solutions doesn’t seem to have their focus in reducing specific fall risk factors like loss of muscle mass or a poor balance. The proposed ICT-based fall prevention solution is based on an existing fall prevention exercise programme specifically designed for the Portuguese population. It takes advantage of the smartphone processing capabilities as well as its built-in inertial sensors to evaluate the movements performed during the execution of specific exercises. Using only a simple smartphone it is possible to provide a friendly and inexpensive solution capable of increasing seniors’ adherence to fall prevention exercises as well as raise their motivation to properly execute the exercises in their home environment.

A Smartphone-Based Fall Risk Assessment Tool: Measuring One Leg Standing, Sit to Stand and Falls Efficacy Scale

V. Guimarães, D. Ribeiro, and L. Rosado

IEEE 15th International Conference on e-Health Networking, Applications Services (Healthcom), 2013, pp. 529–533

Falls are not an inevitable consequence of ageing. Several fall risk factors can be identified and effective fall prevention techniques applied, which offer an opportunity to reduce falls among older persons. In this paper, the smartphone is proposed as an alternative to traditional methods in the assessment of fall risk factors, including decline in balance, reduced lower limb strength and fear of falling. As such, clinical fall risk assessment tests were adapted to the smartphone in order to measure One Leg Standing, Sit to Stand and Falls Efficacy Scale. Experimental results of the system support the feasibility of a reliable phone-based fall predictor, which constitutes an alternative to evaluate fall risk factors in ageing.

Accelerometer-Based Fall Detection for Smartphones

B. Aguiar, T. Rocha, J. Silva, and I. Sousa

IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2014, pp. 1–6

Falls are considered the main cause of fear and loss of independence among the elderly population and are also a major cause of morbidity, disability and health care utilization. In the majority of fall events external support is imperative in order to avoid major consequences.
Therefore, the ability to automatically detect these fall events could help reducing the response time and significantly improve the prognosis of fall victims. This paper presents a unobtrusive smartphone based fall detection system that uses a combination of information derived from machine learning classification applied in a state machine algorithm. The data from the smartphone built-in accelerometer is continuously screened when the phone is in the user’s belt or pocket. Upon the detection of a fall event, the user location is tracked and SMS and email notifications are sent to a set of contacts.
The accuracy of the fall detection algorithm here proposed is 97.5% for the pocket usage and 97.6% for the belt. In conclusion, the proposed solution can reliably detect fall events without disturbing the users with excessive false alarms, presenting also the advantage of not changing the user’s routines, since no additional external sensors are required.

Human Activity Classification with Inertial Sensors

J. Silva, M. Monteiro, and F. Sousa

Stud. Health Technol. Inform., vol. 200, pp. 101–104, 2014

Monitoring human physical activity has become an important research area and is essential to evaluate the degree of functional performance and general level of activity of a person. The discrimination of daily living activities can be implemented with machine learning techniques. A public dataset provided during the European Symposium on Artificial Neural Networks 2013, with time and frequency domain features extracted from raw signals of the smartphone inertial sensors, was used to implement and evaluate an activity classifier. Using a decision tree classifier, an accuracy of 86% was achieved for the classification of walk, climb stairs, stand, sit and lay down. The results obtained suggest that the smartphone’s inertial sensors could be used for an accurate physical activity classification even with real-time requirements.

Monitoring Physical Activity and Energy Expenditure with Smartphones

B. Aguiar, J. Silva, T. Rocha, S. Carneiro, and I. Sousa

IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI), 2014, pp. 664–667

Monitoring physical activity and energy expenditure is important for maintaining adequate activity levels with an impact in health and well-being. This paper presents a smartphone based method for classification of inactive postures and physical activities including the calculation of energy expenditure. The implemented solution considers two different positions for the smartphone, the user’s pocket or belt. The signal from the accelerometer embedded in the smartphone is used to classify the activities resorting to a decision tree classifier. The average accuracy of the classification task for all activities is 99.5% for the pocket usage and 99.4% when the phone is used in the belt. Using the output of the activity classifier we also compute an estimation of the energy expended by the user. The proposed solution is a trustworthy smartphone based activity monitor, classifying the activities of daily living throughout the entire day and allowing to assess the associated energy expenditure without causing any change in the user’s routines.

Phone Based Fall Risk Prediction

V. Guimarães, P. M. Teixeira, M. P. Monteiro, and D. Elias

Wireless Mobile Communication and Healthcare, K. S. Nikita, J. C. Lin, D. I. Fotiadis, and M.-T. A. Waldmeyer, Eds. Springer Berlin Heidelberg, 2012, pp. 135–142

Falls are a major health risk that diminishes the quality of life among older people and increases the health services cost. Reliable and earlier prediction of an increased fall risk is essential to improve its prevention, aiming to avoid the occurrence of falls. In this paper, we propose the use of mobile phones as a platform for developing a fall prediction system by running an inertial sensor based fall prediction algorithm. Experimental results of the system, which we still consider as work in progress, are encouraging making us optimistic regarding the feasibility of a reliable phone-based fall predictor, which can be of great value for older persons and society.