//Logo Image
「世大智科/天才家居」-我們創業囉
PDF Version

作者:劉育瑋 (2014-07-17);推薦:徐業良(2015-12-11)

附註:本文為103學年度元智大學機械工程研究所劉育瑋博士論文「結合軟質活動感知床墊建立以床為核心之遠距居家照護系統」第二章。

Chapter 2. Development and commercialization of the integrated motion sensing mattress, WhizPAD

2.1    Design of the motion sensing mattress, WhizPAD

Figure 2-1 shows the innovation process of WhizPAD. From the development of soft sensing technology to the commercial product of sensing mattress, the design target follows the careful consideration of user preferences and cost. After discussing with home users and nursing staff, the design requirements of WhizPAD are as following.

Ÿ   Functionality: Enabling motion sensing and data transmission;

Ÿ   Usability: Soft, comfortable material and easy to clean;

Ÿ   Production: Low cost and easy to manufacture.

Figure 2-1. The Innovation process of WhizPAD

Based on the above requirements, WhizPAD is designed into a thin mattress pad made of memory foam and conductive textile materials, as shown in Figure 2-2. WhizPAD is a mattress with motion sensing capability using the same material and fabrication process of the bedding manufacturer, so that the mattress is comfortable, flexible in use, easy to install, and low cost. Figure 2-3 shows a possible layout of the sensing areas on the mattress, with three horizontal sensing areas for detecting movements of the upper limbs, hip, and lower limbs. Data of body size from institute of occupational safety and health of Taiwan was used as reference for the layout design. The layout of the sensing areas can be easily adjusted depending on the application.

Figure 2-2. WhizPAD

Possible layout of the mattress 

Figure 2-3. A possible layout of WhizPAD

WhizPAD is in a sandwich structure of two pieces of foam, each 6~10 mm in thickness, on which conductive material is designed in the sensing area, with pieces of conductive foam in between. The working principle of sensing area is simple. If there is no pressure applied to the sensing area, top layer and bottom layer do not contact with each other and results in a broken circuit. Once the sensing area is under pressure, top layer and bottom layer make contact with each other and create a connected circuit. As shown in Figure 2-4, the average resistance of 10 tests of a 20 cm × 20 cm sensing area decreases monotonically with applied pressure in the range of 1,800 ~ 4,300 Pa (the range of pressure caused by the presence of an adult) when measured on surfaces of different hardness. The special elastic foam provided by the bedding manufacture has passed the fatigue test of 30,000 pressure cycles.

Figure 2-4. Relationship between the applied pressure and resistance of sensing units on surfaces of different hardness (blue line: floor, green line: elastic foam, red line: memory foam).

In order to enhance the comfort and decrease the occurrence of bedsores, WhizPAD integrates with the body-shaped memory foam that is atop the sensing layer. The hardness and elasticity of the memory form changes with body temperature, which helps to decrease the stress (<32 mm Hg) applied on the skin. An experiment was conducted to evaluate body pressure distribution measured on WhizPAD. Ten testers, seven males, three females, weighing 58-87 kg were recruited for the experiment. In the first test, testers lay on a standard mattress of a nursing bed for 20 minutes. In the second test, the WhizPAD is put on top of the nursing bed and the testers lay on the WhizPAD for 20 minutes. As shown in Figure 2-5, three Big-Mat sensor sheets (Nitta Corporation) were used to measure the body pressure distribution in the upper limb, hip, and lower limb areas. The average body pressure is 17.2% lower when the WhizPAD is put on top of a standard mattress of a nursing bed. And the highly responsive foam is used in the bottom layer of WhizPAD, so that the sensing performance of WhizPAD is not affected by the type of material of the base mattress on which it is placed. Table 2-1 shows the specifications of the WhizPAD.

Figure 2-5. The body pressure distribution of a 76 kg tester measured on the (a) a standard mattress of nursing home, and (b) WhizPAD using three Big-Mat sensor sheets (Nitta Corporation)

Table 2-1. Specifications of the WhizPAD

Characteristic

Specification

L×W×H

188 cm × 90 cm × 5 cm

Weight

5.94 kg

Major materials

Foam and conductive material

Operational voltage / current

DC 5V / 1 mA

Environment temperature /

humidity

0 ~ 50 °C /

30 to 80%, No condensation

Sensing layer

Sensor type

Piezoresistance

Response time

50/100/500/1000ms

Pressure sensing range

1800 ~ 4300 (N/m2)

Resistance range

3 ~ 1600 (Ohm)

2.2    Development of the BDP

The Gerontechnology Research Center (GRC, http://grc.yzu.edu.tw) of Yuan Ze University has been developing a range of home telehealth applications. The concept of “Decentralized Home Telehealth System (DHTS)” has been proposed. What sets this system different from most others is its focus on a highly decentralized monitoring modality and the portable nature of the system [Hsu et al., 2007].

Figure 2-6 illustrates the information structure of the DHTS. The distributed data server (DDS) is the core unit of DHTS and it has four main functions: receiving data from remote sensors and devices, data logging, data processing and Internet communication. The Internet accessibility of the DDS offers the integratibility to telehealth application and Internet-enabled capabilities. For data request, the DDS can be directly accessed from remote authorized clients using the Internet browsers (e.g., the IE), client App, or even social networking App. This proposed system also provides timely alert reports that respond to emergent events or irregular activities. A centralized database can be optionally connected to DHTS if additional applications are required.

Figure 2-6. The information structure of the Decentralized Home Telehealth System (DHTS) [Hsu et al., 2007]

Based on the structure of the DHTS, there are several advantages over the traditional centralized database structure:

(1)       The scale of the DHTS is much smaller, which makes it economically viable and acceptable to the end-users. A single household can be a running unit of the DHTS. This distributed structure can be adapted if a centralized database is needed.

(2)       Instead of sending the health monitoring data to a centralized database in a home health care provider, health monitoring data are stored within the household. Only authorized caregivers can access the data. Privacy is better protected.

(3)       The route from the sensor to server is much shorter. Data transmission is easier and more reliable. When the Internet communication fails, the local system can still function normally and keep collecting data. Thus data integrity is better preserved.

The WhizPAD is connected to a bedside data processor (BDP) for signal processing, data storage, data transmission and internet connection. The role of the BDP is the same as the distributed data server (DDS) in the Decentralized Home Telehealth System, to be the core of telehealth system.

The Raspberry Pi is used as the core of the BDP for WhizPAD. The Raspberry Pi has a Broadcom BCM2835 system on a chip (SoC), which includes an ARM1176JZF-S 700 MHz processor, VideoCore IV GPU, and is originally shipped with 512 megabytes of RAM. The Raspberry Pi has two USB ports and a 10/100 Ethernet controller, and enables the internet network by using RJ45 Ethernet port or external USB Wi-Fi adapter. An SD card in Raspberry Pi is used for system booting and data storage.

The BDP integrates the Raspberry Pi, 6-channel A/D converter, real-time clocks and ZigBee transmission module. The resistance of a sensing unit of WhizPAD caused by the applied pressure is converted into a voltage signal using a corresponding divided circuit. Through a 10-bit output analog to digital converter, the resolution of pressure signal from a sensing unit is in the range of “0” to “1023”. The sampling rate of the signals from WhizPAD is set at 40 Hz. And the sensing data from the other sensors can be received and applied by the ZigBee transmission module. Details of main capabilities of the BDP are shown in Table 2-2.

  

Figure 2-7. The “Raspberry Pi” is used as the core of the BDP for WhizPAD.

Table 2-2. Detail capabilities of the BDP

Data processing

The ARM1176JZF-S provides fundamental signal processing capability at maximum clock rate of 700MHz. It has 512Mbyte flash program memory.

Analog-to-digital converter and digital input/output (I/O)

The BDP offers 6 analog-to-digital channels, labeled A0 through A5, each of which provides 10 bits of resolution (i.e. 1024 different values). There are 2 digital inputs (Pin Pc0, Pc1) reserved for switches to configure time setting of the real-time clock on BDP.

Micro SD card storage :

BDP uses a micro SD card for data storage, which can be used to store files for serving over the network. Data can be recorded in text files (*.txt) in FAT16 format.

Serial data communications (ZigBee data transmission module):

The BDP provides 1 serial RS-232 channels that are commonly used for communication with a computer, another BDP, or other microcontrollers. In our application, the RS-232 channel connects to the Zigbee microchip for data transmission. The BDP also provides serial I2C connectivity for internal communication between the microcontroller and electronic components.

Internet network communication

An Ethernet controller is used for enabling TCP/UDP connectivity with static IP address and MAC address. The Internet connectivity of the BDP is the most important capability among ordinary embedded system-based devices.

Dividing circuit

It is for converting resistance changes into voltage changes.

ZigBee is a newly developed technology for wireless sensor network (WSN) and a suite of high-level communication protocols using small, low-power digital radios based on the IEEE 802.15.4 standard for wireless home area networks (WHANs). In this study, ZigBee transmission technology is applied in the local wireless data transmission. The ZigBee transceiver module designed for the DDS uses the XBee Series 2 OEM RF module (Digi International) as shown in Figure 2-8.

A typical ZigBee network consists of one “coordinator” and one or more “routers” and/or “end devices”. Figure 2-9 illustrates an example of typical ZigBee PAN network topology. A PAN-ID is required for a coordinator to initiate a valid network that allows other routers and/or end devices with the same PAN-ID to join. When a router or end device joins a PAN, it receives a 16-bit network address from the coordinator and can transmit data to or receive data from other devices in the PAN. Routers and the coordinator can allow other devices to join the PAN, and can assist in sending data through the network to ensure data is routed correctly to the intended recipient device. Note that end devices in a PAN can transmit or receive data but cannot route data from one node to another, nor can they allow devices to join the PAN. End devices must always communicate directly with their parent routers or the coordinator they joined to.

Figure 2-8. The ZigBee transceiver module

Figure 2-9. An example of typical ZigBee PAN network topology

2.3    Development of basic functions of WhizPAD

Given the event algorithms implemented in Atmega644p, the pressure signals collected by WhizPAD can be used to detect on/off bed status, sleep posture and movement in bed. These functions are defined below:

Definition of on/off bed status detection:

The detection of on/off bed on the WhizPAD can be easily recognized by using a simple pressure threshold. In the data acquisition process described above, the average noise of the pressure signals obtained from the BDP is “10”, and the pressure signal of a 30 kg object is about “500”. Therefore “on bed” status is recognized when the sum of pressures signal from all three sensing units is larger than “100”. The definition of on/off bed detection is describing as following:

On bed status: The data of upper limbs + hip + lower limbs > Threshold on-bed

Off bed status: The data of upper limbs + hip + lower limbs < Threshold on-bed

Definition of sleep posture detection:

When lying on the side, the pressure applied by the shoulder is higher than when lying flat. Therefore, the ratio of pressure signals obtained from sensing unit 1 (upper limbs) and sensing unit 2 (hip) defined as R12, is used to determine sleep posture of lying flat and lying side. In a calibration process with 20 subjects, 10 males and 10 females, weighing 40~90 kg, the average R12 of lying flat was 0.32 (σ= 0.15), and the average R12 of lying side was 0.65 (σ= 0.15). Finally a threshold of R12 = 0.6 is used to detect sleep postures of lying normal and lying on the side on the WhizPAD. The definition of sleep posture detection is describing as following:

Normal lying: The data of upper limbs / hip > Threshold posture

Side lying: The data of upper limbs / hip < Threshold posture

Definition of movement in bed detection:

When the status is “on bed”, change in pressure signal in each sensing unit is checked every 1 second to detect whether there is movements on the WhizPAD. As described earlier, the noise of the pressure signals obtained from the BDP is about “10”, and the amplitude of breathing signals is about “40”. Therefore a “movement” is identified if the difference in pressure signals in 1 second from any of the three sensing units is larger than “80”. The definition of movement in bed detection is describing as following:

Current data – data from the previous second > Threshold movement

An experiment was designed to evaluate the performance of the event algorithms. In this experiment, 15 healthy testers, 7 males, 8 females, aged 20-30 years old, weighing 45-98 kg were recruited. Each tester followed a specific procedure: lying flat on the bed, turning to the left side, turning to the right side, then getting off bed. Each motion was maintained for 30 seconds, and the whole procedure was repeated 3 times for each tester (Total case number is 45). Table 2-3 shows the sensitivity and positive predict value (PPV) of WhizPAD. The sensitivity of on-bed detection and off-bed detection are both 1.00; the lying flat and lying side detection in sleep posture detection are 0.79 and 0.92; the movement count detection is 1.00. PPV of on-bed detection and off-bed detection are also both 1.00; the lying flat and lying side detection in sleep posture detection are 0.86 and 0.84; the movement count detection is 0.94. The sensitivity and PPV of recognizing these three events ranges from 0.79 to 1.00 in this experiment. Sleep posture detection has lower sensitivity and PPV.

Table 2-3. The sensitivity and positive predict value (PPV) of WhizPAD system in on/off bed, sleep posture and movement count detections

Base functions

On/off bed

detection

Sleep posture

detection

Movement count detection

Types

On Bed

Off Bed

Lying flat

Lying side

Sensitivity

1.0

1.0

0.79

0.92

1.0

PPV

1.0

1.0

0.86

0.84

0.94

2.4    Respiration rate detection of WhizPAD

WhizPAD also detects respiration rate when a subject is lying on the bed. The main purpose is not to replace existing standard, accurate medical equipment for determining respiration rate, rather, the purpose is to be able to distinguish from a deadweight and a living person lying on the WhizPAD. As shown in Figure 2-10, sensing areas of upper limbs and hip are used to measure the change of body pressure when the subject breathes on the bed. Figure 2-11 shows the signals of physical activities collected by 2 sensing areas of upper limbs and hip of WhizPAD from a 60 kg silica gel model and an 80 kg person. In Figure 2-11(b), the respiration pattern can be seen clearly from signals collected by WhizPAD, while in Figure 2-11(a), the signals obtained from a dead weight put on the bed appear to be background noise.

Figure 2-10. Sensing areas of upper limbs and hip for measurement of respiration

圖片132

Figure 2-11. Signals of physical activities in bed collected by WhizPAD

A procedure of signal processing is proceeded to determine the respiration rate from the pressure signals collected by WhizPAD as shown in Figure 2-12. First the pressure signals are filtered by a 10 points averaging filter. According to the slope, the filtered signals are then transformed into a series of 1 (positive slope) and 0 (negative slope).

In clinical practice, Polysomnography (PSG) is used as the standard equipment for sleep quality evaluation. The BWII PSG from Sleep Virtual is used in this study. It is Type I AASM compliant, composed of 29 channels of parameters, and its max sampling rate is 1000 Hz and signal resolution is 12 bit. Figure 2-13 shows the comparison of respiration signal collected by WhizPAD and by a thorax sensing belt of Polysomnography (PSG). A complete respiration cycle can be extracted from the series of 0/1 data, and the period of the respiration cycle can be calculated. In order to determine the period of respiration, a conservative threshold of period is set between 2.5 and 10 second in this study [Folke et al., 2002]. For example of Figure 2-12c, first respiration was formed by 27 continuous “1” and 35 continuous “0” and, its period is 3.8 second. Finally, the analyzed respiration rate will be output every 60 second.

Figure 2-12. The signal processing of measured respiration of WhizPAD

圖片2rgwgrw

Figure 2-13. The comparison between PSG and WhizPAD

Ten testers, 8 males, 2 females, aged 20-30 years old, weighing 45-90 kg were recruited in an experiment for evaluating the accuracy of respiration rate determined by the WhizPAD. Each tester wore a thorax sensing belt of PSG to measure changes in thorax during respiration. When the experiment started, each tester lay on the WhizPAD and breathed normally for 1 minute. Respiration signal measured can be displayed and stored in the computer for further processing. The whole procedure was repeated five times for each tester (total case number is 50).

The respiration rate output from the WhizPAD was then compared with the integer number of complete respiration cycle detected by PSG. As shown in Table 2-4, the average difference of respiration per minute determined by the WhizPAD is 0.63/min higher than the integer number of complete respiration cycle detected by PSG. Inferential statistical analysis is also used to estimate the difference between WhizPAD and PSG under different confidence levels. Table 2-5 shows the confidence intervals of mean of difference between WhizPAD and PSG, under 90%, 95%, 97.5% and 99% confidence levels (See Appendix A for details of calculating the confidence interval). From the test, we can see that the respiration rate detected by WhizPAD is higher than the integer number of respiration detected by PSG, but the difference is not more than 2.

Table 2-4. The comparison of respiration per minute output by WhizPAD and PSG

Tester

Test

PSG

WhizPAD

Difference

A

1

11.89

11

+0.89

2

10.83

10

+0.83

3

8.79

8

+0.79

4

8.63

8

+0.63

5

10.46e

9

+1.46

B

1

6.97

7

-0.03

2

9.03

8

+1.03

3

10.16

9

+1.16

4

11.83

12

-0.17

5

10.71

10

+0.71

C

1

11.52

10

+1.52

2

11.93

10

+1.93

3

9.95

8

+1.95

4

8.40

8

+0.83

5

7.23

7

+0.23

D

1

6.95

7

-0.05

2

8.02

7

+1.02

3

7.75

7

+0.75

4

7.17

8

-0.83

5

8.93

8

+0.93

E

1

10.57

10

+0.57

2

10.56

10

+0.56

3

11.72

11e

+0.72

4

10.78

9

+1.78

5

11.01

11

+0.01

F

1

6.40

6

+0.40

2

8.58

8

+0.58

3

7.94

8

-0.06

4

6.79

6

+0.79

5

7.86

7

+0.86

G

1

11.71

12

-0.29

2

10.12

10

+0.12

3

11.01

10

+1.01

4

10.49

9

+1.49

5

10.29

10

+0.29

H

1

8.49

8

+0.49

2

9.35

10

-0.65

3

8.89

8

+0.89

4

8.80

9

-0.20

5

9.45

9

+0.45

I

1

11.00

11

+0

2

6.49

7

-0.51

3

10.19

9

+1.19

4

10.63

10

+0.63

5

7.68

7

+0.68

J

1

13.49

13

+0.49

2

12.31

12

+0.31

3

13.01

12

+1.01

4

10.20

10

+0.20

5

10.09

10

+0.09

 

+0.63

Table 2-5. The confidence interval for evaluating the mean of difference between WhizPAD and PSG

Confidence Level

100 ×(1- α/2)%

Lower Bound

Upper Bound

90%

0

1.17

95%

--0.12

1.28

97.5%

-0.22

1.38

99%

-0.33

1.50

2.5    Sleep status recognition of WhizPAD

In sleep, motor activity is reduced in comparison to the waking state, the frequency of all movements decreases with depth of sleep, with progressive decrease in the number of movements from sleep stage I to stage IV [Chokroverty et al., 2003]. Sadeh et al. [1995] used logistic regression analysis for the variables while the sleep/wake classification of PSG acted as the dependent variable in the analysis. Five activity variables were computed for each epoch in the activity signals: original value, mean, standard deviation, number of epochs above a specified activity level, and the natural logarithm. Most actigraphy related devices scored signals using similar algorithms.

Rachwalski et al. [2005] applied a pressure pad to place below a person’s hips, to measure activity in bed. Activity in bed is measured by changes in the pressure measurements. The pressure measurements aggregated into 30-second epochs by averaging the data every 30 seconds. Awakenings (or periods of restlessness) are defined by 3 consecutive minutes of body movements. Choi et al. [2007] designed a bed actigraphy system for distinguishing between sleep and awake. Signals were coded as “1” as the intensity of signals higher than the threshold. If the duration of “1” is longer than 3 seconds in an epoch (30 second length), the epoch is scored as “awake”.

Considering the quantity and tendency of physical activity in bed, Cheng et al. proposed a sleep/awake distinguishing algorithm for the determination of sleep state [2008]. The tendency of activity, weighted activity indices are calculated by considering the activity indices of the past k” minutes with different weighting as shown in Equation 2-1, 2-2 and 2-3. In following equations, the k was set at 5, and the activity indices including “Upper Limb Activity Index (ULAI)”, “Body Activity Index (BAI)” and the “Leg Activity Index (LAI)” represent the length of activities occurred in the past minute.

                  (2-1)

                       (2-2)

                        (2-3)

In the study, WhizPAD can tallies the ULAI, BAI and the LAI for each minute. Figure 2-14 shows a 24 hour record of the ULAI, BAI and LAI of a subject evaluated by WhizPAD. ULAI, LAI and BAI, which represent the duration of activities happened in a minute are quantitative indices for assessing physical activities. These indices can be used for the diagnosis of motor disturbances that are triggered by sleep such as restless legs syndrome (RLS) and periodic limb movements during sleep (PLMS).

Figure 2-14 Activity indices of a subject from 0:00 to 24:00 AM

Eight testers, 6 males, 2 females, aged 20-30 years old, weighing 45-85 kg were recruited in a sleep experiment for evaluating the accuracy of sleep/awake classifying. These testers lay on WhizPAD, wore the brain wave detectors of PSG, and took a nap for 2 hours. The analysis of signals of brain wave collected by PSG can be the reference of sleep depth. In this study, a professional technician (ID: ) helped to score the sleep states and report. The monitoring data of physical activities in bed is also collected by WhizPAD per minute. A total of ____ minutes of sleep data were collected.

To explain the relationship between physical activities and sleep depth, sleep reports from PSG and physical activity data measured by the WhizPAD were collected from 8 subjects. And a sleep value (SV) was defined based on the PSG report and scored by minute. The SV of the minute is coded “0” if PSG recognized two epochs (1 minute) as awake, and “1” as two sleep epochs were detected. If there were only one sleep epoch in the past minute, the SV was coded “0.5”. Regression analyses were used to explain the relationship between physical activities and SV. Regression analyses were used on 500-minute epochs randomly selected from each subject, with weighted activity indices and the fixed k value (k=5) in Equation 2-1, 2-2 and 2-3. Equation 2-4 is the regression equation obtained in this study, in which BAI and LAI were considered for evaluating sleep states with the k value of 5.

                   (2-4)

Figure 2-15 shows a 1 and half hour recording of weighted-BAI, weighted-LAI, and SV calculated with equations 1, 2, and 3. Figure 2-16 shows a 1 and half recording (12:00 to 13:30 PM) of In-Bed Codes and Sleep Codes of the same data in Figure 2-15, as well as the final output “Sleep State.” In this study, the Sleep Code of the past minute is assigned as “1” if SV of this minute calculated by equation 2-4 is higher than 0.5, and “0” if SV is under the threshold limit 0.5.

Figure 2-15 Activity indices of a subject from 0:00 to 24:00 AM

As shown in Figure 2-16, Sleep States were coded “2” (sleep) if both Sleep Code and In-bed Code are “1,” and “0” (Empty Bed) if both Sleep Code and In-Bed Code are “0.” Sleep State “1” (Awake) is recognized as the subject is in bed but stays awake (In-bed Code 1 and Sleep Code 0). According to the output of the WhizPAD in Figure 2-16, this subject fell asleep at the 6th minute. After one and a half hours of sleep, the subject woke up at about the85th minute and left bed at the 95th minute. Figure 2-17 and Figure 2-18 show the sleep states detected by the WhizPAD and PSG of two subjects (Case 1 and 2) in this study. Case 1 is a male subject who slept for one and a half hours but woke many times during the period; Case 2 is a male subject who slept for two hours without awake.

Figure 2-16 In-bed Code, Sleep Code, and Sleep States evaluated by the WhizPAD

Figure 2-17 Sleep states scored via WhizPAD and polysomnographic (Case 1, M, 29 Yrs)

Figure 2-18 Sleep states scored via WhizPAD and polysomnographic (Case 8, M, 21 Yrs)

Table 2-6 shows the comparison results of locating sleep epochs using the WhizPAD and PSG. A one-minute epoch was coded true positive (TP) if the WhizPAD and PSG both classified as sleep epoch, and true-negative (TN) if the WhizPAD and PSG both classified as non-sleep epoch (awake and empty bed). Similarly, a segment was coded false-positive (FP) or false-negative (FN) if only the WhizPAD or only the PSG recognized the state as sleep. Furthermore, in Table 2-6, sensitivity is defined as TP/ (TP+FN), the proportion of sleep epochs correctly identified by the WhizPAD and the total sleep epochs identified by PSG. The positive predictive value (PPV) is defined as TP/ (TP+FP), which refers to the proportion of sleep epochs correctly identified by the WhizPAD and the total sleep epochs identified by the WhizPAD. In the validation test, the sensitivity of locating sleep epochs was 95.6% and the average PPV was 85.1%.

Table 2-6. Comparison results of locating sleep epochs using the WhizPAD and PSG

Case

Sleep Time

TP

FN

FP

Sensitivity

PPV

1

67

55

14

12

79.7%

82.0%

2

116

90

1

26

98.9%

77.5%

3

116

103

2

13

98.0%

88.7%

4

123

96

2

27

97.9%

78.0%

5

113

83

1

30

98.8%

73.4%

6

117

110

4

7

96.4%

94.0%

7

142

126

2

16

98.4%

88.7%

8

112

110

4

2

96.4%

98.2%

Total

906 (min)

773

30

133

95.6%

85.1%

References

Chokroverty S., Hening W. A., Walters A. S., 2003. “Sleep and movement disorders,” Philadelphia: Elsevier Science.

Choi, B.H., Seo, J.W., Choi, J.M., Shin, H.B., Lee, J.Y., Jeong, D.U., Park, K.S., 2007. “Nonconstraining sleep/wake monitoring system using bed actigraphy,” Med Biol Eng Comput, Vol. 45, pp.107–114.

Cheng, C. M., Hsu, Y. L., Young, C. M., 2008. “Development of a portable device for tele-monitoring of physical activities during sleep,” Telemedicine and e-Health, Vol. 14, No.10, pp. 1044-1056.

Folke M., Granstedt F., Hök B., Scheer H., 2002. “Comparative provocation test of respiratory monitoring methods,” Journal of clinical monitoring and computing, Vol. 17, No. 2, pp. 97-103.

Hsu Y. L., Yang C. C., Tsai T. C., Cheng C. M., Wu C. H., 2007. “Development of a decentralized home telehealth monitoring system,” Telemedicine and e-Health, Vol. 13, No.1, pp. 69-78.