//Logo Image
AuthorsChe-Chang Yang (2011-01-20); recommended: Yeh-Liang Hsu (2011-02-21).
Note: This article is the Chapter 8 of Che-Chang Yang’s doctoral thesis “Development of a Home Telehealth System for Telemonitoring Physical Activity and Mobility of the Elderly”.

Chapter 8. Discussion and future work

This thesis presents the development of a system for mobility and functional ability telemonitoring for the elderly based on the decentralized home telehealth system (DHTS). This system incorporates the use of an accelerometry-based wearable motion detector and the home ADL sensors to monitor physical activity and activities of daily living, respectively. The wearable motion detectors achieved real-time activity identification and fall detection, estimation of energy expenditure (EE), and the capability of real-time gait cycle parameters recognition is also demonstrated. The home ADL monitoring can provide the rhythms of daily activities on a long-term basis. The ADLs can be further characterized by means of the ADL features that indicate the activity performances in terms of intensity, frequency and regularity. The detailed information provided by this monitoring system can facilitate the assessment of mobility and functional ability of the elderly people. The summary of this thesis and the future work are briefly introduced in this chapter.

8.1 Discussion

8.1.1 Wearable systems for real-time physical activity monitoring

The wearable motion detector uses accelerometry measurement and a range of signal processing techniques to achieve specific capabilities for the system. The capabilities and design aspects of the wearable motion detector are discussed as follows.

(1)  Hardware and instrumentation

The wearable motion detector is developed for real-time physical activity monitoring. In this device, a capacitive tri-axial accelerometer is used to measure accelerations due to human motions. Capacitive accelerometers have the well-recognized advantages of small form-factors (size), low power consumption, accurate and fast response to motions, which are desirable for wearable systems. The wearable motion detector also utilizes the 2.4GHz ZigBee wireless sensor network (WSN) technology which enables efficient data transmission at low data rate and low power requirement. The advancement in sensor networking enhances the usability and capability in home telehealth applications

(2)  Real-time physical activity identification

For the wearable system using microprocessors, a hierarchical algorithm is developed to identify several basic activities in real-time. Sit-stand postural transitions, lie-sit postural transitions, walking and falls can be identified. Though the preliminary performance test shows good activity identification accuracy, human movements in real living environments may be more complex, and the resulting identification accuracy would be reduced. Despite that, the demonstration of continuous physical activity monitoring using the wearable motion detector at home also shows that the system is suitable and technically feasible for home use. The profile of the identified activities also better reflects the subject daily living and activity status. By implementing the algorithm, the wearable motion detector features real-time capability of activity identification, which is different from most of the existing systems of relevant research. Tele-care and the personal emergency response system (PERS) may be the potential applications utilizing wearable system of such real-time processing capability.

(3)  Estimation of energy expenditure

The estimation of energy expenditure (EE) during physical activity has been widely studied. In this system, a feature-based estimation model is developed for real-time estimating EE. The model output is compared with the oxygen uptake variables, which has been regarded as a gold-standard of EE. Among the 18 activity features extracted from the accelerometry signals, the peak frequency, tilt angles with respect to y- and z-axis, and angular velocity with respect to x-axis are the features posing greater effect on the EE estimation performance. The other features of walking speed, heart rate derived from the GPS and heart rate sensor also has significant effect on the estimation performance. The feature of differential pressure extracted from additional atmospheric pressure sensor also better improves the performance. Therefore, the ability of measuring walking speed and altitude can be considered for designing wearable motion detectors in the future.

(4)  Recognition of gait cycle parameters

The capability of real-time gait cycle parameters recognition using the wearable motion detector was also demonstrated. This trial is expected to enable real-time identification of abnormal gaits, such as shuffling, festinating or freeze of gait (FOG) from Parkinson’s disease patients. The autocorrelation procedure is used to give the temporal gait information of cadence, step regularity, stride regularity and step symmetry. The fast Fourier Transform (FFT) is commonly used to compute the frequency components (power spectrum) of gait patterns. However, it requires advanced computation capability which could only be provided by sophisticated wearable hardware. The autocorrelation is adopted in this study because this method does not require heavy computation capacity and the processing task can be implemented in a microcontroller-based embedded system. From the investigation of gait pattern of the healthy young subjects and the Parkinson’s disease patients, the autocorrelation can discriminate the gait cycle characteristics between the subjects of varied mobility. However, the adverse gaits on the verge of falling or loss of balance were not observed in this study. If possible those adverse gaits data can be very useful for developing advanced gait recognition algorithm.

8.1.2 Home ADL monitoring for the elderly

The other research category of home activity monitoring system is the monitoring of activities of daily living of the elderly living alone at home. Using simple, low-cost and less diverse sensors modality, the passive infrared (PIR) and the current transformer (CT) were adopted in the home ADL sensors. The PIRs can detect active movement related to mobility and the CTs can detect the usage of electrical home appliances of particular interest associated with daily activities. Based on the DHTS, the home ADL monitoring system was installed in a real home environment for long-term monitoring the ADLs of an elder person living alone. The subject was not aware of the system operation and thus this system did not interfere with the daily livings of the elderly subject. The subject did not report any discomfort brought by the system. The ZigBee WSN technology offers reliable ADL data delivery and it also simplifies complex instrument setup in a home environment. Compounding the above advantages, this system can be suitable for home use.

For ADL data analysis, a selection of ADL features is designed in terms of activity intensity/frequency and regularity. These ADL features were also used to analyze the 5-month activity data collected from different locations of the home. These average activity profiles show distinct daily activity rhythms from different locations that relate to specific ADL performances. With the use of the ADL features, the performance in daily activities can be characterized by a set of simple and quantitative estimates. Furthermore, the abnormal events such as post-fall activities and atypical TV usage were recorded during the monitoring period. These events also resulted in significant changes in the ADL features as well as in the activity profile. The detection of atypical events which are largely deviated from its normal rhythm could be important.

8.2 Future work

In this thesis, a home activity monitoring system incorporating wearable motion detectors and home ADL sensors is developed and presented for mobility and functional ability telemonitoring for the elderly. With rapid technological advances, the instrumentation should be adequately improved with the utilization of available technologies. To completely validate the robustness and reliability of the developed system, more trials and tests in home environments, and the target users’ response and acceptance toward this system are required for future development tasks.  

For the wearable motion detectors, a signal processing method has been demonstrated to discriminate normal gaits and Parkinsonian gaits. However, there is still a future development focusing on the realization of an assistive tool for mobility rehabilitation of Parkinson’s disease patients. Featuring the capability in real-time gait pattern recognition, the wearable motion detector will be able to identify gait disorders representing the loss of postural balance, and thus possible falls and fall-related impairment (e.g., hip fractures) can be avoided by instant prompting verbal cues for the patients. This development can benefit PD patients and people with gait disorders by providing a practical tool for assistive rehabilitation. Novel usages and applications of the wearable systems are also possible.

In this thesis, the home ADL monitoring has been demonstrated to be technically feasible, and can provide the information of the rhythms and behaviors of daily activities of the elderly. The ADL performances can also be characterized in terms of quantitative estimates. The functionality of this system can be better enhanced by integrating the application programs and advanced information and communication technologies. These future developments may prompt the potential in commercialization of this system.