«Sensors 2013, 13, 10151-10166; doi:10.3390/s130810151 OPEN ACCESS sensors ISSN 1424-8220 Article Measuring Center of ...»
Sensors 2013, 13, 10151-10166; doi:10.3390/s130810151
Measuring Center of Pressure Signals to Quantify Human
Balance Using Multivariate Multiscale Entropy by Designing
a Force Platform
Cheng-Wei Huang 1, Pei-Der Sue 1, Maysam F. Abbod 2, Bernard C. Jiang 3,4
and Jiann-Shing Shieh 1,4,*
1 Department of Mechanical Engineering, Yuan Ze University, Chung-Li 32003, Taiwan;
E-Mails: firstname.lastname@example.org (C.-W.H.); email@example.com (P.-D.S.) 2 School of Engineering and Design, Brunel University, London UB8 3PH, UK;
E-Mail: firstname.lastname@example.org 3 Department of Industrial Management, National Taiwan University of Science and Technology, Taipei 106, Taiwan; E-Mail: email@example.com 4 Center for Dynamical Biomarkers and Translational Medicine, National Central University, Chung-Li 32001, Taiwan * Author to whom correspondence should be addressed; E-Mail: firstname.lastname@example.org;
Tel.: +886-3-463-8800 (ext. 2470); Fax: +886-3-455-8013.
Received: 3 June 2013; in revised form: 27 July 2013 / Accepted: 1 August 2013 / Published: 8 August 2013 Abstract: To assess the improvement of human body balance, a low cost and portable measuring device of center of pressure (COP), known as center of pressure and complexity monitoring system (CPCMS), has been developed for data logging and analysis.
In order to prove that the system can estimate the different magnitude of different sways in comparison with the commercial Advanced Mechanical Technology Incorporation (AMTI) system, four sway tests have been developed (i.e., eyes open, eyes closed, eyes open with water pad, and eyes closed with water pad) to produce different sway displacements.
Firstly, static and dynamic tests were conducted to investigate the feasibility of the system.
Then, correlation tests of the CPCMS and AMTI systems have been compared with four sway tests. The results are within the acceptable range. Furthermore, multivariate empirical mode decomposition (MEMD) and enhanced multivariate multiscale entropy (MMSE) analysis methods have been used to analyze COP data reported by the CPCMS and compare it with the AMTI system. The improvements of the CPCMS are 35% to 70% (open eyes test) and 60% to 70% (eyes closed test) with and without water pad. The AMTI Sensors 2013, 13 10152 system has shown an improvement of 40% to 80% (open eyes test) and 65% to 75% (closed eyes test). The results indicate that the CPCMS system can achieve similar results to the commercial product so it can determine the balance.
Keywords: center of pressure (COP); multivariate empirical mode decomposition (MEMD);
multivariate multiscale entropy (MMSE)
1. Introduction Postural stability is an important feature that protects people from falls and helps to complete the desired actions, which can be measured by the displacement of the center of pressure (COP) [1,2]. Due to the complicated and nonlinear information signals merging in the brain, the human brain reconstructs the environment from the incoming stream of often ambiguous sensory information and generates unambiguous interpretations of the world to control balance. There is a lot of related research about balance control , however, this study is based on measuring the human COP  signal. It is hoped to establish a balance measurement system to receive COP signals and to distinguish their characteristics, then find the COP signal difference of these multiple sources of sensory information between easy-fall elderly, normal elderly, and normal people. Due to the complicated and nonlinear nature of COP signals, the multivariate empirical mode decomposition (MEMD) would be a good candidate for decompose these signals. After filtering the COP signals, the multivariate multiscale entropy (MMSE) and complexity index (CI) need to be defined to evaluate how good it is for you to control your body in a stable condition. A single value named complexity index (CI) was obtained as the area under the MMSE curve. It gives us an index for assessing the degree of human body balance.
There are many commercial COP measurement systems [such as Advanced Mechanical Technology Incorporation (AMTI), Kistler, CATSYS 2000, etc. [5–7]]. These systems are a PC-based test systems, which communicate with a PC via a serial link, and data are recorded using different sensors. These force plates are specifically designed for use in gait and balance analysis. The systems are valuable for scientific researchers working on diseases, drug development, and monitoring of personnel. However, their prices are usually too high to be affordable for use in homecare systems.
Hence, it is necessary to create a cheaper balance measurement system, with similar performance as commercial COP measurement systems, for preventing falls of elderly at home for the rapid growth of the aging population in the developing (e.g., China) and developed countries (e.g., USA, Europe, Japan).
Furthermore, commercial systems can only collect the COP data and show the locus on screen, but do not provide analysis functions to assess the degree of human body balance. However, it is hard to see any characteristic patterns or indexes from these complicated and highly nonlinear COP data loci.
Fortunately, several non-linear theories have been developed (MEMD, MMSE, CI) which can be used to analyze COP data for assessing the degree of human body balance [2,8–10]. Hence, the main purpose of this study is to design a cheap and portable device, which allows people to use it at home, which can collect and analyze data to assess the degree of human body balance. Therefore, the design procedure for the center of pressure and complexity monitoring system (CPCMS) force platform system is illustrated in the second part. In order to investigate the feasibility of the CPCMS, static, Sensors 2013, 13 10153 dynamic, and correlation tests show were performed in the third part. The fourth section introduces the analysis algorithms used in this study. The empirical mode decomposition (EMD), and MEMD combining with multiscale entropy (MSE) and MMSE using different intrinsic mode functions (IMFs) were compared. The aim is to identify which combination and which IMFs to used that will be accurate and can distinguish the different sways for COP signals. Experimental work is planned for four different sways (i.e., eyes open, eyes closed, eyes open with water pad, and eyes closed with water pad) and compared with a commercial COP measurement system (AMTI). In parts five and six experimental results were analyzed and interpreted.
2. The Design of COP Measurement Device
The structure of the balance measurement device is shown in Figure 1. This device consists of two systems, the pressure measurement platform and the data receiving systems. When the subject stands on the balance measurement device, the pressure measurement platform can receive raw COP signals.
Then the data receiving system can convert the raw COP analog signals into digital signals for use in a computer or display device. A high-resolution data logging system is used with a 16-bit A/D card (i.e., National Instruments (NI) USB-6212) to provide accurate measurement data. The system is named Center of Pressure and Complexity Monitor System (CPCMS).
Figure 1. The center of pressure and complexity monitor system (CPCMS) design flow chart.
2.1. The Pressure Measurement Platform The pressure measurement platform is designed to receive raw balance signal data. Because the load cell signals are very small and have added noise, the signals need to be amplified and filtered.
Therefore, the pressure measurement platform system includes five parts; load cell sensor, Wheatstone bridge, filter circuit, amplifier circuit, and calibrating circuit. The pressure sensing principle of the electronic weight scale is the same as the strain gauge. When the electronic weight scale is under pressure, load cells in the scale will produce deformations. The resistance will change because of the deformation, but the resistance change is difficult to measure, so measurements were based on voltage changes rather than resistance changes. This principle is used as the basis for the measurement of body balance signals. The load cells are embedded in the four corners of the scale. Each load has three signal lines which use a Wheatstone bridge circuit to collect the output voltage changes due to pressure.
Sensors 2013, 13 10154
2.2. Data Receiving System Design
The data logging system uses 16 bits analog to digital (A/D) card to receive the voltage from the pressure measurement platform. After the A/D card converts the voltage signal into a digital signal, the data is saved then analyzed using Matlab. The NI USB-6212 A/D card samples the data at 400 kHz rate, and 16-bit resolution. Four analog input channels were used to receive the signals from the pressure measurement platform, and send the digital signals to a computer through high-speed USB data streams. C programming was used to develop the system which displays and saves the raw data (Figure 2).
2.3. Multivariate Multiscale Entropy Analysis Multivariate Multiscale Entropy (MMSE)  analysis is a method of measuring the complexity of finite length time series. This computational tool can be applied both to physical and physiological data sets, and can be used with a variety of measures of entropy. More details about this analysis method are shown in Section 4.
3. CPCMS Test
3.1. Static Tests
In order to evaluate the static reproducibility, a bowling ball is placed at eight different locations, as shown in Figure 3(b). The CPCMS was calibrated using static state tests. Each location was tested 10 times and the COP location is drawn on a figure. The mean and standard deviation values of each location were also calculated. The results shown in Table 1 indicate the mean of the error ratio for X direction is 7.5% so the maximum error distance for X direction is 0.86 cm (the farthest X coordinate 11.5 cm multiply mean of the error ratio 7.5%), and the mean of the error ratio for Y direction is 11.5% so the maximum error distance for Y direction is 1 cm (the farthest Y coordinate
9.5 cm multiply mean of the error ratio 11.5%).
3.2. Dynamic Tests In order to evaluate the reproducibility of the dynamic state displacement, a dynamic simulating device is designed Figure 4a in order to generate regular dynamic displacement. Driven by high torque motors, a weighed ball rolls on the surface. Therefore, a regular circle of displacement is generated as shown in Figure 4b, and the circle is plotted using radius as the mean of total excursion plus the standard deviation of all excursions (mean + SD) . The mean and standard deviation values of the radius after ten times testing were calculated. The results are shown in Table 2. The real radius for simulating device is 10 cm, and the error ratio for totally 10 times tests is 4% equal to
0.4 cm. The dynamic test error ratio 4% is better than the static test error ratio of 10% for CPCMS.
3.3. Correlation Analysis Test The correlation of the two different signals can be determined using a cross-correlation function.
If two signals are similar to each other, the correlation value will be close to 1. Conversely, if two signals are not similar, the correlation value will go to −1 (Figure 5). The maximum correlation values were also identified. The raw COP data measured with the AMTI and CPCMS were aligned to the same start point before calculating the correlation value.
The correlation results are shown in Table 3. It is found that the correlation values are close to 1.
The correlation value of sway 4 type is the smallest one of the four sways, but still close to 1. It means that the COP sway data measured by the two systems are similar.
Sensors 2013, 13 10157
3.4. Intraclass Correlation Coefficient Test Intraclass correlation coefficient (ICC) has been frequently used for reliability of measurements .
ICC values of bigger than 0.75 represent “excellent reliability” and values between 0.4 and 0.75 represent “fair to good reliability” . Static and dynamic tests data were used to calculate the ICC.
In static tests the ICC for top right sensor X is 0.66 and Y is 0.99, bottom right sensor X is 0.97 and Y is 0.99, top left sensor X is 0.87 and Y is 0.92, and bottom left sensor X is 0.74 and Y is 0.98. In the dynamic test the ICC is 0.88. From the results it can be seen that the CPCMS has excellent reliability except for the top right sensor (X = 0.66) and bottom left sensor (X = 0.74).
4. Analysis Algorithms
The COP measurement data were analyzed using empirical mode decomposition (EMD) and MEMD for detrending the data. Since the data is contaminated with noise, it needs to be filtered prior to analysis. The EMD method is an iterative signal processing algorithm which decomposes the intrinsic components from signals by iterative sifting processes. Then, MSE is used to evaluate the body COP. MEMD is a newly modified method originally from EMD, primarily is to create white noise, of the length as the original signal, and then feed the original signal and white noise in different channels. MEMD is not only suitable for dealing with multichannel signals, but also solve the problem of mode mixing addition of white Gaussian noise to different channels (N-A MEMD) .
MSE analysis is a new method of measuring the complexity of finite length time series.