//Logo Image
Authors: Che-Chang Yang, Yeh-Liang Hsu. (201-06-02); recommended: Yeh-Liang Hsu (2012-08-27).
Note: This paper is published at Journal of Clinical Gerontology and Geriatrics, 2012, v. 3, n. 2, pp. 1-8, doi:10.1016/j.jcgg.2012.06.002.

Remote monitoring and assessment of daily activities in the home environment

Abstract

Background/Purpose: Quantitative analysis of daily activities measured by home monitoring systems can be helpful to objectively assess the health-related living behaviors and functional ability of the older adults. Advances in sensors and telecommunication technologies have prompted the concept and development of implementing diverse sensors and intelligent algorithms to monitor the human-environment interactions. This paper presents the remote monitoring and assessment of daily activities of older adults living alone at home, assuming that comprehensive profiles of daily activities at home can be captured by using simple and low-cost sensors in a less diverse modality.

Methods: A passive home monitoring system with minimal set of sensing modalities, namely human infrared and electrical current, was developed and used for continuous and unobtrusive monitoring of daily activities of the subject for over 6 months. Four movement detectors were deployed in different indoor locations to detect active movements, and an appliance usage detector detected the television use. A set of activity features that measure the intensity, regularity and abnormalities of activity patterns is defined and demonstrated to quantify the characteristics and rhythms of daily activities of the subject.

Results: Different rhythms of daily activities can be estimated from different locations at home, and distinct behaviors were shown between weekdays and holidays. Unusual activities have been detected by the system, too. The system setup did not require modification of home furnishings that could obstruct the subject’s daily living or cause any discomfort, displeasure to the subject.

Conclusion: This study suggests that daily activities of an older adult living alone at home can be measured by means of low-cost sensors in a less-diverse sensor modality, and the daily rhythms can be quantified with a simple estimation method. The activity features developed in this study are built into a home telehealth system for telecare applications.

Keyword: daily activity; older adults; telecare; remote monitoring.

1. Introduction

Activities of daily living (ADLs) refer to several daily tasks which are required for personal self-care and independent living, such as eating, dressing or bathing.1 Lawton et al. also defined the instrumental ADLs (I-ADLs) as the activities that require interaction with objects or instruments for self-care or communications with people, such as the use of telephone or basic home appliances.2 The ADL performance directly reflects an individual’s living independency. Ageing process is an expected cause of limiting ADL performance that changes from advanced or moderate ADLs to a lower, basic ADL level.3

The performance of daily activities has been widely adopted in clinical and research fields to evaluate the level of disability, or functional status of elder people. For example, the ADL Scales,4-6 the I-ADL Scales,2 Barthel Index (BI)7 and the Functional Independence Measure (FIM) scales8 have been developed to assess the functional ability. Traditional ADL assessment methods usually rely on self reports, diaries, questionnaires or subjective judgments by clinical or specialized personnel. Technologies have the potentials to assist ADL measurements in an unobtrusive way without disturbing the daily life of the older adults. Long-term activity profiles of the older adults monitored at home environment can provide additional comprehensive information related to the living behaviors, and thus their functional ability can be better determined objectively. In 1993, the research group in the University of New South Wales (UNSW) initiated the development of a cost-effective sensor-based remote monitoring approach to identifying changes of health status with sensors and simple activity measures.9,10 In their proposed systems, a telemedicine platform was used to enable effective, automatic and continuous data transmission and collection, as well as to prompt timely response and care intervention.

Advances in sensors and telecommunication technologies have prompted the concept and development of “smart homes” that are achieved by implementing diverse sensors and intelligent algorithms to monitor the human-environment interactions. For healthcare purposes, such smart homes or home monitoring systems can collect health-related data to build a profile of health and functional status of elder people.11 In such smart homes, sophisticated instruments and sensors may be used. However, home activities can also be monitored by using a range of simple and low-cost sensors distributed in a home environment.12 The passive infrared sensor (PIR) is the most common sensor to detect human occupancy or active movements within specific ranges in a space. Mechanical, magnetic or photoelectrical switches can also be used to detect location transfers at home.13 Estimation of energy expenditure using PIRs was also studied, though the preliminary outcome is not acceptable.14 Sensors can also be used to monitor I-ADLs. The ADLife developed by Tunstall [http://www.tunstall.co.uk] is a telecare solution for the elderly people. The system can detect electrical appliance usages (e.g., oven, refrigerators) to provide more detailed information on the elder persons’ I-ADLs within their homes. A study in the monitoring of home electrical appliances usages of elder people living alone showed that daily and nocturnal activities can be differentiated.15 ADLs detection and classification of the usage of electrical appliances from power line impulses were also presented.16

Quantitative analysis of daily activities measured by home monitoring systems can be helpful to objectively assess the health-related living behaviors and functional ability of the older adults. 17 This paper presents the remote monitoring and assessment of daily activities of an older adult living alone at home. It is assumed in this study that comprehensive profiles of daily activities at home can be captured by using simple and low-cost sensors in a less diverse modality. Thus instead of using complex and multiple types of sensors, in this study a home activity monitoring system based on only human infrared sensors (PIR) and electrical current sensors (current transformer, CT) was installed in the residence of an older adult who lived alone. The movement detectors which utilize PIRs were deployed in different indoor locations at home to detect active movements. An appliance usage detector using a CT detected the television use. The only two types of sensors were selected because they are deemed adequate for monitoring home activities related to daily living rhythm. This approach not only simplifies the complexity in instrumentation and the sensor fitting to the home environment, but also provides a unified basis for data analysis. Six-month activity data were collected through the continuous and unobtrusive monitoring. A set of activity features that measure the intensity, regularity and abnormalities of activity patterns is defined and demonstrated to quantify the characteristics and rhythms of daily activities of the subject, namely, active time ratio Ractive-time, activity rate Ract, daily activity rate Ract-day, coefficient of variance of daily activities CVact, correlation coefficient of activity profile r. Unusual activities can be detected by the system, too. These activity features are built into a home telehealth system for telecare applications.

This paper is organized as follows. The design of the monitoring system and the data collection were described in Section 2.1 and 2.2. Section 2.3 described the use of simple activity features for quantitative analysis. The analysis result from the subject was demonstrated in Section 3, and Finally Section 4 concludes this study.

2. Methods

2.1 Instrumentation

Figure 1 shows the structure of the home activity monitoring system developed in this study. This system structure is based on the Decentralized Home-Telehealth System previously developed by the authors.18 In this system, the environment variables related to human daily activities can be measured and detected by means of the home activity detectors. The home activity detectors as shown in Figure 2 are microcontroller-based devices (PIC18LF6722, Microchip Co), and are equipped with ZigBee RF modules (XBee Series 2 OEM RF module, Digi International) to enable wireless sensor networking via 2.4GHz radio band. The home activity detectors basically support versatile and multiple sensor connectivity for specific measurement purposes. In this study only two types of sensors were selected: passive infrared sensors (PIR) and current transformers (CT). The home activity detectors with PIR (movement detectors) as shown in Figure 2 (a) detect noticeable changes in infrared intensity due to human movements within the detection cone of the PIR. The “ON” states (active movements exist) and “OFF” states (inactive, or no movement) can be identified cyclically. The movement detectors transmit ON-state signals to the distributed data server (DDS) once human movements are sensed. The movement detectors can also measure room humidity and temperature with on-board sensors (SHT75, Sensirion AG). The humidity/temperature variables are also transmitted whenever the signal transmission is enabled.

As shown in Figure 2 (b), the other type of home activity detector using the CT is the appliance usage detector. It measures the AC current consumed by the connected home electrical appliances. A threshold for the CT sensor output is set to distinguish the ON (in use) and OFF states (not in use) of the connected electrical appliances. Similar to the movement detectors, the appliance usage detector transmits the ON-state signals to the DDS. Note that the home activity detectors cyclically detect the sensor status every 6 seconds, and the DDS records data every 10 minutes. As a result, the DDS may receive 100 ON state counts in a 10-minute interval if the home activity detectors are continuously triggered for 10 minutes.

The distributed data server (Figure 2 (c)) is an embedded system primarily consists of the same PIC microcontroller and ZigBee RF module. An Ethernet controller (RTL8019AS, RealTek) is also used to enable Internet communication. The DDS records the ON-state counts wirelessly received from the home activity detectors and the data are stored in an MMC memory card. Application programs or the Internet browsers (e.g., the IE) can access and retrieve the data in the DDS.

Fig. 1. The structure of the home activity monitoring system

Fig. 2. The home activity detectors: (a) movement detector; (b) appliance usage detector; (c) the distributed data server

2.2 Subject and the residence environment for data collection

To collect home daily activities, 5 home activity detectors, including 4 movement detectors and one appliance usage detector, were installed in a residence of an older adult (female, 75yr) in this study. The 4 movement detectors were installed in the kitchen, bedroom, and bathroom and by the doorway, respectively, where the subject frequently had her daily activities. Figure 3 shows the setup of the movement detectors. The appliance usage detector interconnects an AC wall outlet and a television to detect the television usage. Television usage is monitored because watching television is one of the major activities of the subject at home. All the home activity detectors were mains-powered and the data can be wirelessly transmitted to a DDS via the 2.4GHz ZigBee WSN protocol in real-time. Informed consent was obtained from the subject, although during the monitoring period, the subject was not aware of the system operation, and the setup did not require modification in home furnishings or home facilities. Thus the system setup did not obstruct, change or interfere with the subject’s daily living routines. No video or audio recording was used during the monitoring period so that the subject’s privacy can be highly preserved and protected.

Fig. 3. The setup of the home activity detectors in the subject’s residence

2.3 The use of simple estimates for daily activity analysis

A selection of simple estimates (activity features) is used to quantify the daily activities in terms of the activity frequency and the activity regularity. Figure 4 shows an example of the activity data collected someday in the kitchen. As the data was recorded every 10 minutes, there are totally 144 time periods (or epochs) in a day. The activity counts in each epoch were shown chronologically in the above of Figure 4. The maximum activity count collected in one epoch is 100. The same data can be re-arranged in the frequency order by ranking the activity counts in the epochs, as shown in the below of Figure 4. The activity features in terms of activity frequency can thus be derived from the frequency-ranked activity data.

Fig. 4. The example of one-day activity data in chronological order (above) and in the frequency order (below)

Feature 1: active time ratio

People may perform their daily activities at varied time periods in a day. The first focus of interest is how frequently the subject performs activities in a daily scale. As shown in the frequency-ranked activities in Figure 5,  is the activity count at the ith epoch. The epochs which have activity counts () are called “active epochs”, and the other epochs without activity counts () are called “inactive epochs”. As a result, the total epochs can be divided into the “active period (Tact)” and “inactive period (Tina)”.

The “active time ratio” is a measure to indicate how frequently the subject performs activities over the entire day. The active time ratio  is defined as the ratio of the number of active epochs (Tact) to overall epochs (Tact+Tina), as expressed in Equation (1). For example, the number of Tact in Figure 5 is 47 and the total number of epochs is 144. This corresponds to  32.6%. A high active time ratio generally indicates a more frequent activity profile, regardless of the activity intensity in the active epochs.

                                                                             (1)

Fig. 5. The active period and the inactive period in a frequency-ranked activity profile

Feature 2: activity rate (or usage rate for appliances)

In addition to the active time ratio, it is also important to have the estimates which regard the intensity of activities. The “activity rate” and “daily activity rate” are defined for this purpose. The activity rate measures the intensity of activity over the active period, while the daily activity rate measures the intensity of activity over the entire day.

The activity rate (, or the usage rate for appliances) is calculated according to Equation (2), which is the total activity counts in the active period Tact divided by the maximum number of the active epochs (i.e., ×100). Note that in Equation (2), if there is no activity in a day, both the sum of  and  are zero. Therefore the activity rate here is defined as zero.

                                                                     (2)

For the same example shown in Figure 5, the total activity count is 528 from the 47 active epochs. As a result, the activity rate is 11.2%, which means that the subject has activities in 11.2% of the time in an active epoch in average. A higher activity rate indicates that the intensity of the activity is higher during the active epochs.

Feature 3: daily activity rate (or daily usage rate for appliance usage)

Similar to Feature 2, the daily activity rate (or daily usage rate for appliances)  generally shows the intensity of the activities the subject performs in a whole day. For the sample data in Figure 5, the daily activity rate is 3.67%, which is an indicator of the intensity of activity of the day.

                                                                              (3)

Feature 4: coefficient of variance of daily activities (or coefficient of variance of daily appliance usage)

People may perform activities of varied frequency at different time periods of the day, which results in varied activity patterns. The coefficient of variance of the activity counts distributing over a whole day () defined in Equation (4) can provide an estimate of how uniformly the subject performs activities in a day. For the example data in as Figure 4, the coefficient of variance is 2.28.

                                                            (4)

Feature 5: correlation coefficient of activity profile (or correlation coefficient of appliance usage profile)

It has been reported that elderly people tend to have stable lifestyle [Franco et al., 2008]. Hence, the regularity of daily activities compared with the long-term profile can be an important estimate of whether the subject follows his/her own regular activity rhythms (behavior patterns). The correlation coefficient r is a common statistical measure of the interdependence of two or more variables. Therefore, the correlation coefficient is used to compare a daily profile to the long-term average profile. The correlation coefficient ranges from -1 to 1, and a higher value in the correlation coefficient of two profiles means a more similar or correlated trend between the two profiles. For example, the correlation coefficient of the one-day activity profile shown in Figure 4 and the 6-months average profile as shown in Figure 6 is 0.17.

Fig. 6. The long-term profile of daily activities in the kitchen

3. Results

3.1 The long-term activity profiles

Table 1 lists the activity features analyzed from the 6-month activity data in the kitchen, bedroom, doorway and the bathroom, and the long-term average profiles are also graphically shown in Figure 7. The activities in the kitchen have greater daily activity rate, active time ratio and activity rate. These activity features indicate a relatively greater activity frequency in the kitchen among the other three locations at home. The smallest coefficient of variance of daily activities and highest correlation coefficient of activity profile are also found in the kitchen activities, indicating a more regular activity rhythm among the others. On the contrary, the activities by the doorway show the least correlation coefficient of activity profile and highest coefficient of variance of daily activities among the other three locations. These activity features show that the activities by the doorway are least regular, which is also expected. According to the activity features, the subject had relatively intensive activities in the kitchen, and this may imply that the subject still preserves a good performance level in the I-ADLs.

Fig. 7. The long-term average activity profiles in (a) kitchen, (b) bedroom, (c) doorway, and (d) bathroom of the elderly subject’s residence

Table 1. Comparison of the activity features computed from the 6-month average activity profile in the four locations at home

ADL features

Kitchen

bedroom

doorway

bathroom

Daily activity rate

2.19%

0.11%

0.19%

0.93%

Active time ratio

0.19

0.06

0.09

0.10

Activity rate

11.73%

1.79%

1.89%

9.29%

Coefficient of variance of daily activities

2.89

4.73

4.55

4.72

Correlation coefficient of activity profile

0.35

0.22

0.20

0.26

Figure 8 shows the long-term average profile of TV usage, and its activity features are listed in Table 2. The most intensive TV usages are at 9:40 in the morning and 19:20 in the evening. From the high correlation coefficient of daily TV usage 0.74, the subject showed regular behaviors in TV usage.

Fig. 8. The long-term average profile of TV usage

Table 2. Comparison of the 6-month average profiles of TV usage

 

Activity features

TV usage

Daily usage rate

49.03%

Active time ratio

0.50

Usage rate

97.28%

Coefficient of variance of daily appliance usage

1.06

Correlation coefficient of appliance usage profile

0.74

Figure 9 shows the long-term average activity profiles on weekdays and holidays in the kitchen, bedroom, doorway, and bathroom. Moderate differences between the profiles on weekdays and holidays can be observed from this figure. Table 3 lists the detailed ADL features from Figure 9. In general, the subject’s daily activities between weekdays and holidays remain similar. This shows there is no significant difference regarding the subject’s activity rhythm between weekdays and holidays.

Fig. 9. The long-term average activity profiles between weekdays and holidays in (a) kitchen, (b) bedroom, (c) doorway, and (d) bathroom of the elderly subject

Table 3. Comparison of the activity features computed from the 6-month average activity profile in the four locations at home (W: weekdays; H: holidays)

Activity features

Kitchen

bedroom

doorway

bathroom

W

H

W

H

W

H

W

H

Daily activity rate

2.07%

2.30%

0.10%

0.12%

0.17%

0.21%

0.85%

11.00%

Active time ratio

0.19

0.19

0.06

0.07

0.09

0.10

0.10

0.10

Activity rate

11.11%

12.35%

1.79%

1.80%

1.79%

2.00%

8.73%

9.85%

Coefficient of variance of daily activities

2.87

2.92

4.98

4.49

4.38

4.45

4.84

4.61

Correlation coefficient of activity profile

0.35

0.35

0.21

0.22

0.17

0.23

0.26

0.25

Figure 10 shows the long-term average TV usage on weekdays and holidays, and Table 4 lists its activity features. The higher daily usage rate 55.29% on weekdays, which means that the TV is ON more than half of the day, indicates that watching TV is the major activity of the subject at home. Compared with the weekday profile, the holiday profile has lower daily usage rate, active time ratio and usage rate. This shows the subject used TV more on weekdays than on holidays. The greater coefficient of variance of daily appliance usage and the lower correlation coefficient of appliance usage in the holiday profile also indicate a less regular TV usage than weekdays. By interviewing the subject, it has been confirmed that the subject usually watches TV news and programs to help handle finances (stock market and transactions) in the morning during weekdays. This is considered as a major cause of the different performance in the activity features and the significant changes observed from the two profiles.

Fig. 10. The long-term average profiles of TV usage on weekdays and holidays

Table 4. Comparison of the average profiles of daily TV usage on weekdays and holidays

Activity features

Weekdays

Holidays

Daily usage rate

55.29%

42.76%

Active time ratio

0.56

0.44

Usage rate

98.85%

95.72%

Coefficient of variance of daily appliance usage

0.92

1.19

Correlation coefficient of appliance usage profile

0.82

0.67

3.2 Unusual activities

Unusual activities have been detected during the monitoring period. For example, Figure 11 shows the activities in the bedroom collected on April 29, 2010. Unusual activities were found during 3:10 to 3:50am as the subject actually fell accidentally in her bedroom in the early morning and could not get up. Table 5 shows the activity features of the collected activities on that day and the 6-month long-term activity average. The unusual activities cause higher activity rate, daily activity rate and active time ratio. Though the coefficients of variance of daily activities from both profiles are similar, the correlation coefficient of daily activities on that day is 0.01, which is largely lower than the counterpart 0.22 from the 6-month long-term average.

Fig. 11. The one-day activity data with unusual activities in the bedroom

Table 5. The activity features from the activities collected in the bedroom on April 29, and the 6-month long-term activity average

Date

Activity features

April 29

Long-term average

Daily activity rate

0.7%

0.1%

Active time ratio

0.11

0.06

Activity rate

5.94%

1.8%

Coefficient of variance of daily activities

4.87

4.73

Correlation coefficient of daily activities

0.01

0.22

Figure 12 shows unusual TV usage from 2:00 to 2:40 detected on July 12, 2010. It was later confirmed that the subject was watching TV in the early morning on that day because she could not fall asleep. As listed in Table 6, the correlation coefficient of daily appliance usage on that day (r=0.59) is lower than that from the 6-month long-term average (r=0.74). The other ADL features roughly remain similar.

Fig. 12. The TV usage data collected on July 12, 2010.

Table 6. Comparison of activity features in the TV usages on on-day profile and 6-month the long-term average profile

Activity features

July 12

Long-term average

Daily usage rate

44.8%

49.0%

Active time ratio

0.47

0.50

Usage rate

96.2%

97.3%

Coefficient of variance of daily appliance usage

1.11

1.1

Correlation coefficient of daily appliance usage

0.59

0.74

4. Discussion

In this paper, a home activity monitoring system is presented. This system was installed in a residence of a subject living alone to monitor the subject’s home daily activities. In this system only two types of home activity detectors, the movement detectors using the PIRs and an appliance usage detector using the CT, were used because this approach is considered to not only simplify the complexity in instrumentation and the sensor fitting to the home environment, but also provide a unified basis for data analysis. The movement detector senses apparent human movements in different locations at home. The appliance usage detector was connected to a television to measure the television use. During the monitoring period the subject was not aware of the operation of the system. The system setup did not require modification of home furnishings that could obstruct the subject’s daily living or cause any discomfort, displeasure to the subject. The wireless data transmission using ZigBee WSN in a home environment is also reliable. Although the particular living characteristics and rhythms were obtained from the monitoring system in such a single-subject study, home visitors or co-habitants (e.g., families, home nurses) living together at home cannot be distinguished by those simple sensors. This technical issue limits the system usability if the system is considered to be used in family residences or for community-dwelling subjects.

A set of activity features in terms of activity frequency and regularity was adopted to provide quantitative estimates regarding the home activity characteristics by analyzing the 6-month activity data collected by the system installed in a residence of an older adult (female, 75yr) in this study. These simple estimates can indicate the characteristics and the long-term living rhythms of daily activities of the subject living alone at home. Unusual activities have been detected by the system, too. The results from this study also suggest that the use of less-diverse and simple sensors may be sufficient for remote home activity monitoring. The older adult is actually the mother of one of the authors of the paper, and she was fully aware of the test and agreed with it. The older adult and the son actually welcomed the test because the sensors are gathering information not only for healthcare purposes. Knowing the living rhythms of the older adult also enriches the content of interaction and communication between the older adult and the son.

The activity features developed in this study are built into the home telehealth system. The monitoring system and the analysis method presented in this paper can be a cost-effective approach to evaluating functional status of the older adults for personal home healthcare applications.

References

1.     Katz S, Ford AB, Moskowitz RW, Jackson BA., Jaffe MW. Studies of illness in the aged: The index of ADL, a standardized measure of biological and psychosocial function, JAMA, 1963:185:914-919.

2.     Lawton MP, Brody EM. Assessment of older people: Self-maintaining and instrumental activities of daily living, The Gerontologist, 1969:9:179-186.

3.     Pennauthor A, Magham R, Contreras L.R. Dowling W. Daily living activities in older adults: Part I- a review of physical activity and dietary intake assessment methods, International Journal of Industrial Ergonomics, 2003:32:389-404.

4.     Katz S. Downs T.D. Cash H.R. Grotz R.C. Progress in development of the index of ADL, The Gerontologist, 1970:20-30.

5.     Katz S. Assessment self-maintenance: Activities of daily living, mobility and instrumental activities of daily living”, Journal of the American Geriatrics Society, 1983:31:721-727.

6.     Katz S. Akpom C. A measure of primary sociobiological functions, International Journal of Health Service, 1976:6:493-508.

7.     Mahoney FI. Barthel D. Functional evaluation: the Barthel Index, Maryland State Med Journal, 1965:14:56-61.

8.     Tinetti ME, Performance-oriented assessment of mobility problems in elderly patients, J Am Geriatr Soc. 1986:34:119–126.

9.     Celler BG. Hesketh T. Earnshaw W. Ilsar E. An instrumentation system for the remote monitoring of changes in functional status of the elderly at home, In: Proceedings of the 16th Annual International Conference of the IEEE EMBS, 1994:2:908-909.

10.  Celler B.G. Earnshaw W. Ilsar, E.D. Betbeder-Matibet L. Harris M.F. Clark R. Hesketh T. Lovell NH. Remote monitoring of health status if the elderly at home. A multidisciplinary project on aging at the University of New South Wales, International Journal of Bio-medical Computing, 1995:40:147-155.

11.  Noury N. Virone G. Barralon P. Ye J. Rialle V. Demongeot J. New trends in health smart homes, In: Proceedings of 5th International Workshop on Enterprise Networking and Computing in Healthcare Industry. 2003:118–127.

12.  Ogawa M. Togawa T. Monitoring daily activities and behaviors at home by using brief sensors, In: 1st Annual International IEEE-EMBS Special Topic Conference on Microtechnologies in Medicine & Biology, 2002.

13.  Suzuki R. Otake S. Izutsu T. Yoshida M. Iwaya T. Monitoring daily living activities of elderly people in a nursing home using an infrared motion detection system, Telemedicine and e-health, 2006:12:146-155.

14.  Kaushik A. Celler BG. Use of infrared sensors for estimation of energy expenditure by elderly people living alone at home, In: Proceedings of the 28th IEEE EMBS Annual International Conference, 2006.

15.  Franco GC. Gallay F. Berenguer M. Christine M. Couturier P. Non-invasive monitoring of the activities of daily living of elderly people at home- a pilot study of the usage of domestic appliances, J Telemed Telecare, 2008:14:231-235.

16.  Berenguer M. Giordani M. Giraud F. Noury N. Automatic detection of activities of daily living from detecting and classifying electrical events on the residential power line, In: Proceedings of the 10th IEEE International Conference on e-Health Networking, Applications and Service, 2008.

17.  Suzuki R. Ogawa M. Otake S. Izutsu T. Tobimatsu Y. Iwaya T. Izumi S-I. 2006. Rhythm of daily living and detection of atypical days for elderly people living alone as determined with a monitoring system, J Telemed Telecare, 2006:12:208-214.

18.  Hsu, YL. Yang CC. Tsai, TC. Cheng CM. Wu, CH. Development of a decentralized home telehealth monitoring system, Telemedicine and e-Health, 2007:13:69-78.