novel wearable sensors for biomechanical … · web viewnovel wearable sensors for biomechanical...

E

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2943487, IEEE Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Novel Wearable Sensors for Biomechanical Movement Monitoring Based on

Electromagnetic Sensing TechniquesYuedong Xie*, Mingyang Lu, Wuliang Yin, Senior Member, IEEE, Hanyang Xu*, Shuang Zhu, Jiawei

Tang, Liming Chen, Qiaoye Ran, Yining Zhang, Zhigang Qu

Abstract—For many diseases, treatment is more effective in the early stage of a disease; hence detection of the early signs of a disease is highly significant. Biomechanical motion can be such an early indicator. This paper proposes a novel method for monitoring biomechanical motion based on electromagnetic sensing techniques. The proposed method has the advantages of being non-invasive, easy to perform, of low cost, and highly effective. Theoretical models are set up to model the sensor responses and wearable sensor systems are designed. Experiments are performed to monitor various biomechanical movements, including eye blinking frequency, finger/wrist bending level and frequency. From experiments, both the fast blinking behaviorwith an average frequency of ～1.1 Hz and the slow blinkingbehavior with an average frequency of 0.4 Hz can be monitored; various finger-bending status are identified, such as the fastfinger-bending with an average frequency of ～1.5 Hz and theslow finger-bending with an average frequency of ～1/6 Hz. Bothsimulations and measurement results indicate that the proposed electromagnetic sensing method can be used for biomechanical movement detection and the system has the advantages of being easy to put on / take off, and without the need for direct contact between sensors and human body.

Index Terms—Biomechanical monitoring, Electromagnetic, Analytical solutions, Wearable Sensors.

I. INTRODUCTION

ARLY prevention, early detection, and early treatment are key topics in medical field [1]. Medical diagnosis is the

process to exam the symptoms and medical signs and increasingly people have realised the importance of regular

medical examination [2]. A medical sign is an indication for medical conditions and is important for medical diagnosis [3]. For example, abnormal eye blinking can be used as a clue for

Yuedong Xie and Zhigang Qu are with the College of Electronic Information and Automation, Tianjin University of Science and Technology, 1038 Dagunan Road, Tianjin, 300222, P. R. China. Meanwhile, Yuedong Xie is with Department of Physics, University of Warwick, Coventry CV4 7AL, UK (corresponding e-mail: [email protected]).

Wuliang Yin, Mingyang Lu, Hanyang Xu, Shuang Zhu, Jiawei Tang, Liming Chen, Qiaoye Ran and Yining Zhang are with School of Electrical and Electronic Engineering, University of Manchester, Manchester, M13 9PL, UK (corresponding e-mail: [email protected]).

dry eye syndrome, burning mouth syndrome, progressive supranuclear palsy, and even autism [4-7]. Abnormal motor behaviour, such as tremor and rigidity, could be a medical sign of alcohol intoxication, hypokinesia, hyperthyroidism, hepatic disease, and even parkinsonian syndrome, etc. [8-13]. Since treatment is more effective in the early stage of diseases, detection of the early sign of these abnormal biomechanical behaviours is highly significant, hence portable and convenient detection techniques are vital.

On the other hand, some biomechanical or in general biomedical signals can be used as a mean to communicate between human and electrical devices. Human-machine interfaces (HMIs), a communication method between a human and an external device to turn a virtual thought into realistic action, have drawn increasing attention among researchers. HMIs, which utilizes biomechanical / biomedical signals to express emotions or control external devices, have advantages of “hands-free” [14, 15]. And hence, HMIs are especially helpful for patients suffering from motor disorders. Accurate detection of biomechanical / biomedical signals is the base for the next stage, such as expression and control. Frequently used biomedical signals include electroencephalogram (EEG), electrooculogram (EOG), electrocorticogram (ECG), and electromyogram (EMG) signals [16-23]. All of these biomedical signals have their specific applications and have their individual advantages and disadvantages. Briefly, ECG is a highly informative and diagnostic method, and easy to perform. EEG is a low cost and readily repeatable method but suffers from low spatial resolution and poor signal-noise ratio (SNR). EMG is to detect the electric potential by muscle cells but commonly requires the needle electrodes to be inserted into the muscle tissue and suffers from disturbance.

In this paper, a novel method for biomechanical movement detection is proposed. The proposed method is based on electromagnetic (EM) sensing techniques to detect biomechanical movements, such as eye blinking frequency (eye-blink rate detection is useful for fatigue determination [24, 25], wrist/finger bending level and frequency (bending monitoring is useful for tremor and rigidity determination). More specifically, by attaching the EM sensor to daily wearable articles (such as false eyelashes and gloves), the relative mechanical movement between eyelids/joints can be monitored through monitoring the relative movement between the

∬




attached EM sensors, which leads to a change in mutual impedance /inductance signal between the EM sensors. Although EM sensors have been widely used in industries, they are also increasingly being used biomedical field [26-28]. The proposed methodology and the developed devices have the benefit of being non-invasive, easy to perform, of low cost and effective. Several methods can be used for biomechanical movement monitoring as well, such as infrared radiation (IR) illumination method, scleral search coil method, and optical fibre sensor, etc. Scleral search coil method is one of the precise measurement techniques but suffers from direct contact between contact lens and corneas due to its invasive nature. It is not possible to wear contact lens with coils on for a long time, and hence scleral search coil is only limited to short-term monitoring eye movement. However, using EM sensing to monitor eye movement, due to its non-contact nature, makes long-term monitoring possible [29]. IR illumination is significantly affected by the external light source, which makes it not very stable compared to the proposed EM method [30]. EOG is a non-invasive and inexpensive method but the measurement accuracy strongly depends on skin conditions and luminous, etc. Hence it suffers from a lack of accuracy at the extremes [20]. For bending movement, optical fibre sensor can be used, but it is easily affected by environmental parameters, such as thermal and acoustic vibration interference [31].

Detection principle and related theoretical models are introduced in Section II. Experiments are carried out to validate the effectiveness of the proposed method as an alternative

Fig. 1. The schematic diagram with two co-axially symmetric coils placed above a layered conductor.

The proposed method is based on monitoring the change in mutual impedance/inductance from the EM sensors due to the variation of their relative spatial position. Dodd and Deeds have presented the analytical solutions to the mutual impedance calculation for two circular coils (excitation coil and detection coil) as shown in [35]. Two co-axially symmetric models were introduced in [35], from which the model with coils placed above layered conductor planes were used in this work. The governing equations for electromagnetic sensing phenomena are shown in equations (5) to (7), from which the Equation (5) shows the vector potential generated by a driving coil, Equation(6) presents the induced voltage calculation of the receiving coil, and Equation (7) shows the mutual impedance calculation between the driving coil and the receiving coil.

method for biomechanical signal detection shown in Section III. ∇2𝑨𝑨 = −𝜇𝜇𝒊𝒊 + 𝜇𝜇𝜇𝜇 𝜕𝜕𝑨𝑨 + 𝜇𝜇𝜀𝜀 𝜕𝜕2𝑨𝑨 1(5)

Finally, conclusions and discussion are presented. 𝒕𝒕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕2 + 𝜇𝜇∇( ) × (∇ × 𝑨𝑨)𝑽𝑽𝒓𝒓 = 𝑗𝑗2𝜋𝜋𝜋𝜋𝑛𝑛′ 𝑟𝑟𝑨𝑨(𝑟𝑟, 𝑧𝑧) 𝑑𝑑𝑟𝑟𝑑𝑑𝑧𝑧 (6)𝑆𝑆′

II. EM DETECTION PRINCIPLE, MODELS AND SIMULATIONS 𝒁𝒁 = 𝑽𝑽𝒓𝒓 𝑰𝑰𝒕𝒕(7)

A. EM Sensing BackgroundMaxwell equations, which include Gauss’s law for electric

fields, Gauss’s law for magnetic fields, Faraday’s law, and Ampere-Maxwell law, are a series of partial differential equations to describe the electromagnetic sensing related phenomena [32]. Gauss’s law for electric fields (Equation 1) relates the electrostatic field to the charge distribution that generates it [32, 33]. Gauss’s law for magnetic fields (Equation2) states the total magnetic flux passing through any closed surface is zero [34]. Faraday’s law (Equation 3) describes that the time varying magnetic field generates an electric field [32]. Ampere-Maxwell law (Equation 4), known as Ampere’s law with Maxwell’s addition as well, states that the magnetic field can be generated either by the electric current or the time varying electric fields [32].

where 𝑨𝑨 is the magnetic vector potential, 𝜇𝜇, 𝜇𝜇 and 𝜀𝜀 are the permeability, conductivity and permittivity of the material respectively, 𝒊𝒊𝒕𝒕 is the current density applied to the driving coil,𝜔𝜔 is the angular frequency of the applied alternating current, 𝑽𝑽𝒓𝒓is the induced voltage on the receiving coil, 𝑆𝑆′ and 𝑛𝑛′ are the cross section and the turn of the receiving coil respectively, 𝑰𝑰𝒕𝒕 is the current applied to the driving coil, and 𝒁𝒁 is the mutual impedance between two coils.

Two co-axially symmetric circular coils (symmetry with z direction) placed above a layered conductor are shown in Fig. 1, from which ‘T’ denotes the cross-section of the driving coil and ‘R’ denotes the cross-section of the receiving coil. The final analytical solution for the impedance calculation is shown in Equation (8), which indicates that the mutual impedance is related to the location, dimension and turn of the coil, the∇ ∙ 𝐄𝐄 = 𝜌𝜌

(1) property of the test material, and the driving frequency, etc.𝜀𝜀0∇ ∙ 𝐁𝐁 = 0 (2)𝒁𝒁 = ��𝑗𝜋𝜋𝜇𝜇𝜋𝜋𝑛𝑛𝑛𝑛′𝑟𝑟

∫∞ 1 𝐼𝐼(𝑟𝑟2, 𝑟𝑟1)𝐼𝐼(𝑟𝑟2, 𝑟𝑟1)𝑙𝑙� 2−𝑙𝑙1 𝑟𝑟�� 2−𝑟𝑟1 𝑙𝑙�� 2−𝑙𝑙1 𝑟𝑟�� 2−𝑟𝑟1 � 0 𝛼𝛼5 𝜕𝜕 𝜕𝜕 𝑟𝑟 𝑟𝑟∇ × 𝐄𝐄 = ∂𝐁𝐁

(3) 𝑟𝑟 𝑟𝑟 𝑟𝑟 𝑟𝑟 𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑡𝑡

𝜇𝜇




∂t 2� (𝑙𝑙2 − 𝑙𝑙1) + 𝛼𝛼−1 exp� [−𝛼𝛼(𝑙𝑙2 − 𝑙𝑙1)] − exp[−𝛼𝛼(𝑙𝑙2 − 𝑙𝑙2)] +∇ × 𝐁𝐁 = μ (𝐉𝐉 + 𝜀𝜀 ∂𝐄𝐄 (4) 𝑟𝑟 𝑟𝑟 𝜕𝜕 𝑟𝑟 𝜕𝜕 𝑟𝑟0 0 ∂t) exp[−𝛼𝛼(𝑙𝑙2 − 𝑙𝑙1)] − exp[−𝛼𝛼(𝑙𝑙1 − 𝑙𝑙1)] + {exp[−𝛼𝛼(𝑙𝑙2 +where 𝐄𝐄 is the electric field, 𝜌𝜌 is the charge density, 𝜀𝜀 is the 2 𝑟𝑟 𝜕𝜕 2 1 𝑟𝑟 𝜕𝜕 1 2 𝑟𝑟 1

electric permittivity of free space, 0 𝑙𝑙𝜕𝜕 )] − exp[−𝛼𝛼(𝑙𝑙𝑟𝑟 + 𝑙𝑙𝜕𝜕 )] − exp[−𝛼𝛼(𝑙𝑙𝑟𝑟 + 𝑙𝑙𝜕𝜕 )] + exp[−𝛼𝛼(𝑙𝑙𝑟𝑟 +𝐁𝐁 is the magnetic field, μ0 is 𝑙𝑙1)]} �( 𝛼 𝛼 + 𝛼𝛼 1 ) ( 𝛼𝛼 1 − 𝛼𝛼 2 ) + � ( 𝛼 𝛼 − 𝛼 𝛼 1 ) � ( 𝛼𝛼 1 + 𝛼𝛼 2 ) e x p ( 2 𝛼 𝛼 1𝑐 𝑐 ) (8)

the permeability of free space and 𝐉𝐉 is the electric current 𝜕𝜕 �� 𝑑𝑑𝛼𝛼(𝛼𝛼−𝛼𝛼 )(𝛼𝛼 −𝛼𝛼 )+�(𝛼𝛼+𝛼𝛼 )�(𝛼𝛼 +𝛼𝛼 ) exp(2𝛼𝛼 𝑐𝑐)density. 1 1 2 1 1 2 1




Fig. 2. The schematic of the simulation setup (co-axially movement between EM sensors).

𝛼𝛼𝑖𝑖 = �𝛼𝛼2 + 𝑗𝑗𝜔𝜔𝜇𝜇𝜇𝜇𝑖𝑖 (9) where 𝑙𝑙1, 𝑙𝑙2, 𝑟𝑟1 and 𝑟𝑟2 define the location and the dimension𝜕𝜕 𝜕𝜕 𝜕𝜕 𝜕𝜕of the driving coil, 𝑙𝑙1, 𝑙𝑙2, 𝑟𝑟1 and 𝑟𝑟2 define the location and the Fig. 3. For co-axial movement, the mutual impedance signal at various gap𝑟𝑟 𝑟𝑟 𝑟𝑟 𝑟𝑟 ′ distances between EM coils.

dimension of the receiving coil, 𝑛𝑛 and 𝑛𝑛 are the turn number ofthe driving coil and the receiving coil respectively, 𝑟𝑟 ̅ is the mean radius of the coil, 𝛼𝛼 is the integrated factor, 𝑐𝑐 is the thickness of the 1st layer conductor, 𝜇𝜇𝑖𝑖 means the conductivity of the ith conductor.

B. Biomechanical Movement ModelingThis study aims to detect the biomechanical movement,

including eye blinking frequency, finger bending level and frequency, wrist bending level and frequency. The separate EM sensors are attached to the wearable articles, such as false eyelashes and gloves, etc., and the related biomechanical signals are detected through monitoring the relative movement between separate EM sensors. To validate the feasibility of the study, some simulations based on the analytical models were carried out.The first simulation study is to model co-axial movement between two EM coils; this simulation study is to verify that the relative movement between two circular coils can generate varies mutual impedance signal, which is the foundation for the proposed method using EM sensors to monitoring biomechanical signal. The modelling setup is shown in Fig. 2; two identical coils, one for excitation and the other one for detection, are placed co-axial with a varying centre-to-centre distance. Fix one of the circular coils, and move another one gradually, and calculate the mutual impedance at each movement step. The two circular coils used have identical parameters; the outer radius is 0.25 mm, the inner radius is 0.1575 mm, and the height is 1 mm; the coil parameters in modelling are the same with those used in experiments. The centre-to-centre distance between these two coils is originally1.1 mm, and increases to 11.1 with a step of 0.1 mm. After that, the centre-to-centre distance decreases to 1.1 mm with a step of0.1 mm. This progress was repeated several times. Since the height of the coils is 1mm, hence the gap between Tx and Rx is from 0.1 mm to 10 mm with a step of 0.1 mm. At each step, the mutual impedance is recorded. The excitation coil carries an alternating current with an amplitude of 1 A and the frequency used is 100 kHz.

For the case without conductor included ( 𝛼𝛼 = 𝛼𝛼1 = 𝛼𝛼2 ), Equation (8) becomes,𝒁𝒁 =

Fig. 4. (a) Eye blinking behaviour – the example of relative movement between EM sensor is non-co-axial. Three status during blinking, including fully-opened, half-opened and fully closed; (b) The schematic of the simulation setup (non-coaxial movement between EM sensors); the red block and blue block denote the excitation coil and the detection coil respectively, the green line states the moving path, originally from 0° to 90°.

Based on Equation (10), the mutual impedance of EM coils in air can be obtained. The calculated mutual impedance versus varying centre-to-centre distance is shown in Fig. 3. For this simulation setup, the mutual impedance signal is mainly changed within the gap of 2 mm between the transmitter and the receiver, beyond which the mutual impedance signal fluctuates around 0. From Fig. 3, the mutual impedance signal arrives to peak when the gap between the EM sensors is smallest, which is 0.1 mm in this work. On the other hand, the amplitude of the mutual impedance decays along the centre-to-centre distance between EM coils. In addition, by counting the peak number, the frequency of the EM sensors moved can be obtained. Therefore, the mutual impedance can be used as an indicator for the relative movement detection.

For biomechanical movement behaviours, such as eye blinking, finger bending, etc., the relative movement between eye lids (Fig. 4 (a)) or joints is slightly complicated, instead of the simple co-axial movement. Another simulation study was carried out to validate the relative movement along an arc-path will have an effect on the mutual impedance signals as well. The moving path is shown in Fig. 4 (b); the excitation coil (Tx) is fixed, and the detection coil (Rx) is moved along one quarter circle, which means the detection coil is moved from 0° to 90° if defining the position of two coils co-axial is 90°. The other setup parameters, including the excitation frequency and driven current amplitude, etc., are the same with the one used in the co-axial simulation. The induced voltage is obtained from the integral of the vector potential over the volume of the whole��𝑗𝜋𝜋𝜇𝜇𝜋𝜋𝑛𝑛𝑛 𝑛′𝑟𝑟 ∫∞ 1 𝐼𝐼(𝑟𝑟2, 𝑟𝑟1)𝐼𝐼(𝑟𝑟2, 𝑟𝑟1) (2(𝑙𝑙2 −




receiver, and the impedance between the transmitter and�𝑙𝑙2−𝑙𝑙1��𝑟𝑟2−𝑟𝑟1��𝑙𝑙2−𝑙𝑙1��𝑟𝑟2−𝑟𝑟1� 0 𝛼𝛼5 𝜕𝜕 𝜕𝜕 𝑟𝑟 𝑟𝑟 𝑟𝑟 receiver is computed from the induced voltage on the receiver

𝑟𝑟 𝑟𝑟 𝑟𝑟 𝑟𝑟 𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑡𝑡𝑙𝑙1) + 𝛼𝛼−1{exp[−𝛼𝛼(𝑙𝑙2 − 𝑙𝑙1)] − exp[−𝛼𝛼(𝑙𝑙2 − 𝑙𝑙2)] + coil when excited by the current in the transmitter.𝑟𝑟 𝜕𝜕 𝑟𝑟 𝜕𝜕 𝑟𝑟exp[−𝛼𝛼(𝑙𝑙2 − 𝑙𝑙1)] − exp[−𝛼𝛼(𝑙𝑙1 − 𝑙𝑙1)]})𝑑𝑑𝛼𝛼 (10)𝑟𝑟 𝜕𝜕 𝑟𝑟 𝜕𝜕




Non-coaxial movement25

20

15

10

5

1

0.8

0.6

0.4

0.2

0

Co-alxial Movement1

0.8

0.6

0.4

0.2

0

Arc-path MovementA

mpl

itude

(

)

Nor

mal

ised

am

plitu

de o

f Z N

orm

alis

ed a

mpl

itude

of

Z

(a)




(b)




0 1 2 3 4 5 6 7 8 9 10Gap (mm)

0 20 40 60 80 100Moving positions (deg)

00 20 40 60 80 100

Angle (degree)Fig. 5. For arc-path movement, the mutual impedance signal at various angles between EM coils.

Fig. 6. Customised high-speed EM instrument.The calculation results of mutual impedance for arc-path

movement are shown in Fig. 5. The mutual signal is mainly changed when the detection coil is moved between 70° to 90°. These figures indicate, even though the EM sensors are not moved co-axially, the mutual impedance signal follows the trend that the shortest distance between the transmitter and the receiver, the signal is largest. And hence, either the co-axial movement or the arc-path movement, the mutual impedance can be used as an inductor for biomechanical signal monitoring. One point should be noted is that, in real applications, the EM sensors cannot be moved exactly along co-axial path or arc-path, but these simulations results validate the feasibility that the EM sensing techniques can be used as an alternative method for biomechanical movement detection.

III. EXPERIMENTS

Experiments were carried out to show the application of the EM sensors for biomechanical signal detection. The instrument used is a customised high-speed EM testing instrument based on FPGA (Field Programmable Gate Array), which is able to operate from 5 kHz to 2 MHz, perform digital demodulation at the rate of 100 k/second and features an Ethernet communication to PC as shown in Fig. 6. The excitation coil was driven by an alternating current ~48 mA rms. For the detection coil, the signal was amplified by a PGA (Programmable Gain Amplifier) with a specific gain, which was 4 in this work, and then fed into an analogue to digital converter (ADC). The amplified voltages on the pickup coil were then acquired and recorded by a host PC. The instrument has been demonstrated to provide an SNR (Signal to Noise Ratio) of > 90dB.

A. Co-Axial Movement and Arc Movement ExperimentsInitial experiments were carried out to validate the

simulation carried out in Section II.B. For co-axial movement,

Fig. 7. The measured mutual impedance signal at different movement behaviours between EM coils. (a). co-axial movement; (b). arc-path movement.

Fig. 8. (a) Wearable EM system includes a glass frame, an EM sensor (two coils), false eyelash and connecting cable; (b) A picture with the EM sensors being worn.the transmitter is fixed, and the receiver is moved co-axially. The centre-to-centre distance between the transmitter and the receiver is from 0.5 mm to 5 mm with a step of 0.5 mm and from 5 mm to 10 mm with a step of 1 mm. The measured data for co-axial movements is shown in Fig. 7(a); the curve of the measured mutual impedance signal is normalised by the maximum value among the recorded amplitude values. It can be seen that the value of the mutual impedance signal is the largest when the centre-to-centre distance is the smallest and it decreases with the increase of the distance, which has a similar trend as simulations. For the movement along an arch-path, the sensor was moved from 0° to 90° with a step of ~8°. The measured normalised data are presented in Fig. 7(b). The experimental curve has the same trend with the simulated curve– the signal is largest when the two EM sensors are closest.

B. Monitoring Eye Blinking BehaviorThe designed wearable EM sensor for blinking monitoring

is shown in Fig. 9(a), which consists of a wearable glass, two wearable false eyelashes with EM sensors attached on it, and one Ethernet cable to connect to the EM instrument. With such a low-cost wearable design, the EM sensing for eye blinking monitoring can be easily performed. These two EM coils have identical parameters; the outer radius, the inner radius and the height of the coil are 0.25 mm, 0.1575 mm, and 1 mm respectively. The thickness of the wire of the coil is 0.03mm, and the turn of the coil is 100. Fig. 9(b) presents a photo when the EM system is worn. One coil is attached to the upper eyelid through a false eyelash, and the other coil is attached to the lower eyelid through another false eyelash. With such design, the false eyelash is directly contact with the eyelid but there is no direct contact between coils and human body. The relative movement between the upper and lower eyelids results in the relative movement between the attached coils which forms the EM sensor.




Fig. 9. The measured EM response with the designed wearable EM system for eye blinking monitoring.

Fig. 10. The designed wearable EM system for finger bending monitoring.The experiment is to monitor the alternating progress of fast

blinking and slow blinking; fast blinking for ~5 seconds, and then slow blinking for ~25 seconds. The whole sampling period is ~1 minute. The measured data is shown in Fig. 9, the moment of eye closing during blinking has a peak signal since the coils (EM sensor) are closest. Therefore, fast blinking has denser peak signals while slow blinking has looser peak signals. In addition, the amplitude of the fast blinking signal is larger than that of the slow blinking signal, which is due to the fact that the fast blinking behavior involves in a stronger muscle movement than the slow blinking behavior; in other words, the coils are closer under fast blinking. From 10s to 30s, there are 8 blinks, which means the slow blinking has an average frequency of 0.4 Hz. Similarly, from 36.5s to 40s, there are 15 blinks, which means the fast blinking has an average frequency of ~1.1 Hz. Apparently, by setting an appropriate threshold for the blink frequency/blink amplitude, fast and slow blinking can be distinguished. For different people, a baseline can be established with a preliminary experiment to determine the specific threshold.

C. Monitoring Finger Bending BehaviorThe reason for monitoring the finger bending behavior is

due to the fact that increasing number of people is suffering from carpal tunnel syndrome. An EM sensor has been designed to be wearable for finger bending monitoring as shown in Fig. 10; two coils are attached on a rubber glove at the position of the finger joint of the index finger. The reason for choosing this finger is that index finger is used very often in daily lives, such as clicking mouse.

Fig. 10 presents four finger-bending status, non-bending, slight bending, middle-level bending and full bending. At different bending status, the distance between the EM coils is different. And hence, the mutual impedance signal at different bending status should be different. The experimental setup for the finger bending monitoring is to monitor the alternating progress of fast finger bending and slow finger bending:

bending the finger slowly for ~25 seconds, then bending the finger quickly for ~10 seconds, holding for ~10 seconds, and quickly bending for ~5 second, and holding for ~10 seconds. The whole recording period is ~1 minute. The recorded data for monitoring finger bending is shown in Fig. 11. The signal fluctuates around 0 during full bending and arrives to a peak at the moment of non-bending. From this figure, during the first 30 seconds, the finger was bent slowly for 5 times, and hence the average frequency of the slow bending is roughly ~1/6 Hz. During 31s to 41s, the finger was bent quickly for 15 times, corresponding to an average frequency of ~1.5 Hz. One point should be noted is, these experiments were carried out with keeping each time’s finger bending consistent as much as possible, in other words, the finger was being bent within the same range. From Fig. 11, the peaks of the recorded signals are around 2×10-3 Ω, except the signal in the blue circle because it is difficult to keep each time’s finger bending exactly consistent.

Another experiment was carried out with each time’s finger bending randomly, i.e., the bending range of each time’s finger bending was random. The whole recording period for such random bending is ~90 seconds; the measured amplitude of mutual impedance signal versus time is shown in Fig. 12, from which the fast finger bending occurs at 1s to 12s and the slow finger bending occurs at 12s to 40s, followed by a holding period for 10 seconds, and finally is a slow finger bending from 61s to 92s. Apparently, with such random finger bending, the peaks of the mutual impedance signal are random as well, which is different from the results of each time’s consistent bending – the peaks are roughly the same. Therefore, not only the bending frequency, but also the bending level can be monitored with the designed wearable EM system.




Fig. 11. The measured EM response with the designed wearable EM system for finger bending monitoring; – keeping each time’s finger bending consistent as much as possible.

Fig. 12. The measured EM response with the designed wearable EM system for finger bending monitoring– each time’s bending randomly.

Fig. 13. The wrist bending diagram with the designed wearable EM system.

D. Monitoring Wrist Bending BehaviorSimilar to finger bending behaviour, the wrist bending

behaviour was measured as well. The schematic of the designed wearable EM system for wrist bending monitoring is shown in Fig. 13, from which the EM coils are attached on a glove at the position of the wrist joint. Fig. 13 (a) denotes the non-bending status, (b) denotes the upward bending and (c) denotes the downward bending. Different bending status are with varies relative positions of EM coils, which means the mutual impedance signals are different.

The experimental setup is to monitor the alternating progress of non-bending and upward/downward bending: non-bending for ~5 seconds, and then bending up for ~5 seconds, and so on. The whole recording period is ~35 seconds, and the recorded mutual impedance signal versus time is shown in Fig. 14 (a). There are four peaks around 0.008 Ω, corresponding to 4 times upward bending. Therefore, the amplitude of the mutual impedance signal can be used as the indicator for bending frequency measurement. Another experiment for wrist bending monitoring with the wrist downward bending was carried out, and the results are shown in Fig. 14 (b). The results show 5 times downward bending within ~55 seconds.

The experiment with the combination of the wrist upward bending and the wrist downward bending was carried out. The experimental data are shown in Fig. 15. From the recorded amplitude of mutual impedance signal, the wrist upward bending, non-bending, and the wrist downward bending can be distinguished clearly through defining an appropriate amplitude

Fig. 14. The wrist bending monitoring with the designed wearable EM system;(a) bending up; (b) bending down.

Fig. 15. The wrist bending monitoring with the designed wearable EM system – the combination of upward bending and downward bending.threshold. There are 5 times upward bending and 5 times downward bending respectively. If the non-bending status is defined as the balance status, the upward bending and the downward bending have opposite directions. Ideally, the amplitude of upward bending should same with that of the downward bending if the bending angles of upward bending and downward bending are the same. But in real applications, the amplitude of upward bending and downward bending are different, possibly because the EM sensors are not exactly attached at the centre of the wrist joint or the upward bending range and the downward bending range are not exactly the same. Despite this, through defining an appropriate amplitude threshold, the wrist bending level and the bending frequency can still be measured.

IV. CONCLUSION AND DISCUSSION




Medical diagnose is a hot topic, both for research and clinical applications. Early detection of the biomedical sign allows medical diagnose and the treatment to be more effective. This paper explores EM sensing to detect biomechanical movement, including eye blinking, finger bending and wrist bending. The wearable EM system based on EM sensors and wearable articles are designed, which makes it easy to perform and eliminate the need for direct contact between sensors and human body.

Theoretical models were set up to validate the feasibility of the proposed method. Both the co-axial movement behavior and the movement along an arc-path were studied, and the results indicate the distance between EM coils has an effect on the amplitude of the mutual impedance signal. More specifically, the closest EM coils corresponds to the largest amplitude of the mutual impedance signal.

Experiments are divided into four parts. The first part is to validate the simulations, and the results exhibit the same trend with simulations. The second experiment is to monitor eye blinking frequency. The designed wearable EM system for eye blinking monitoring is based on the EM sensor formed by two EM coils attached on the false eyelashes. The measured results indicate that the fast blinking and slow blinking behaviors can be clearly distinguished with the proposed method, and the measured blinking frequency is exactly the same with the real situations. In addition, the fast blinking results in a larger amplitude than that of the slow blinking because the fast blinking involves a stronger tissue movement compared to slow blinking. The third and fourth experiments are for finger/wrist bending monitoring. Measured results indicate that both the bending level and the bending frequency can be recorded in real time.

Overall, this paper proposes a novel method for biomechanical movement detection based on EM sensing techniques, and wearable EM systems are developed with the benefit of low-cost and easy to perform etc. Both simulations and measurement results indicate that the proposed electromagnetic sensing method can be used for biomechanical movement detection and the system has the advantages of being easy to put on / take off, and without the need for direct contact between sensors and human body. Future work will focus on more robust ways of attaching the EM sensors to false eyelashes, and more sophisticated algorithms for defining the threshold automatically for different patients. In addition, other biomechanical signal monitoring, such as the detection of pulse, heartbeat and eyerolling, will be attempted by combining the proposed electromagnetic sensing method and the technique of micro-electro-mechanical system as another future work.

REFERENCES

[1] C. Levkov, G. Michov, R. Ivanov, and I. Daskalov, "Subtraction of 50 Hz interference from the electrocardiogram," Medical and Biological Engineering and Computing, vol. 22, pp. 371-373, Jul. 1984.

[2] M. C. Peterson, J. H. Holbrook, D. Von Hales, N. Smith, and L. V. Staker, "Contributions of the history, physical examination, and laboratory investigation in making medical diagnoses," Western Journal of Medicine, vol. 156, p. 163, Feb. 1992.

[3] V. L. Malcarne, I. Hansdottir, A. McKinney, R. Upchurch, H. L. Greenbergs, G. H. Henstorf, et al., "Medical signs and symptoms associated with disability, pain, and psychosocial adjustment in systemic sclerosis," The Journal of rheumatology, vol. 34, pp. 359-367, Feb. 2007.

[4] T. Schlote, G. Kadner, and N. Freudenthaler, "Marked reduction and distinct patterns of eye blinking in patients with moderately dry eyes during video display terminal use," Graefe's archive for clinical and experimental ophthalmology, vol. 242, pp. 306-312, Apr. 2004.

[5] L. L. Sears, P. R. Finn, and J. E. Steinmetz, "Abnormal classical eye-blink conditioning in autism," Journal of autism and developmental disorders, vol. 24, pp. 737-751, Dec. 1994.

[6] S. K. Jääskeläinen, H. Forssell, and O. Tenovuo, "Abnormalities of the blink reflex in burning mouth syndrome," Pain, vol. 73, pp. 455-460, Dec. 1997.

[7] L. I. Golbe, F. E. Lepore, and P. H. Davis, "Eyelid movement abnormalities in progressive supranuclear palsy," Movement disorders: official journal of the Movement Disorder Society, vol. 4, pp. 297-302, Apr. 1989.

[8] M. Alpert, B. Angrist, F. Diamond, and S. Gershon, "Comparison of ditran intoxication and acute alcohol psychoses," in Origin and Mechanisms of Hallucinations, ed: Springer, 1970, pp. 245-260.

[9] C. J. Hunker, J. H. Abbs, and S. M. Barlow, "The relationship between parkinsonian rigidity and hypokinesia in the orofacial system: a quantitative analysis," Neurology, vol. 32, pp. 749-749, Jul. 1982.

[10] C. Trivalle, J. Doucet, P. Chassagne, I. Landrin, N. Kadri, J. F. Menard, et al., "Differences in the signs and symptoms of hyperthyroidism in older and younger patients," Journal of the American Geriatrics Society, vol. 44, pp. 50-53, Jun. 1996.

[11] C. Cooper, V. G. H. Evidente, J. G. Hentz, C. H. Adler, J. N. Caviness, and K. Gwinn-Hardy, "The effect of temperature on hand function in patients with tremor," Journal of Hand Therapy, vol. 13, pp. 276-288, Jan. 2000.

[12] J. Shahed and J. Jankovic, "Exploring the relationship between essential tremor and Parkinson's disease," Parkinsonism & related disorders, vol. 13, pp. 67-76, Dec. 2007.

[13] J. J. Geraghty, J. Jankovic, and W. J. Zetusky, "Association between essential tremor and Parkinson's disease," Annals of Neurology: Official Journal of the American Neurological Association and the Child Neurology Society, vol. 17, pp. 329-333, Apr. 1985.

[14] R. Vaidyanathan, B. Chung, L. Gupta, H. Kook, S. Kota, and J. D. West, "Tongue-movement communication and control concept for hands-free human–machine interfaces," IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, vol. 37, pp. 533-546, Jul. 2007.

[15] W. Lai and H. Huosheng, "Towards multimodal human-machine interface for hands-free control: A survey," Techn. Ber., School of Computer Science & Electronic Engineering, University of Essex, Jul. 2011.

[16] L. Wei, H. Hu, and Y. Zhang, "Fusing EMG and visual data for hands-free control of an intelligent wheelchair," International Journal of Humanoid Robotics, vol. 8, pp. 707-724, Dec. 2011.

[17] L. Wei and H. Hu, "Multi-modality—EMG and Visual Based Hands-Free Control of an Intelligent Wheelchair," in International Conference on Intelligent Robotics and Applications, 2010, pp. 659-670.

[18] Z. Lv, L. Feng, H. Li, and S. Feng, "Hand-free motion interaction on google glass," in SIGGRAPH Asia 2014 Mobile Graphics and Interactive Applications, 2014, p. 21.

[19] H. W. Goo and D. H. Yang, "Coronary artery visibility in free-breathing young children with congenital heart disease on cardiac 64-slice CT: dual-source ECG-triggered sequential scan vs. single-source non-ECG-synchronized spiral scan," Pediatric radiology, vol. 40, pp. 1670-1680, Oct. 2010.

[20] H. S. Dhillon, R. Singla, N. S. Rekhi, and R. Jha, "EOG and EMG based virtual keyboard: A brain-computer interface," in 2009 2nd IEEE International Conference on Computer Science and Information Technology, 2009, pp. 259-262.

[21] M. Dhuliawala, J. Lee, J. Shimizu, A. Bulling, K. Kunze, T. Starner , et al., "Smooth eye movement interaction using EOG glasses," in Proceedings of the 18th ACM International Conference on Multimodal Interaction, 2016, pp. 307-311.




[22] C. Neuper, G. R. Müller-Putz, R. Scherer, and G. Pfurtscheller, "Motor imagery and EEG-based control of spelling devices and neuroprostheses," Progress in brain research, vol. 159, pp. 393-409, Jun. 2006.

[23] A. Campbell, T. Choudhury, S. Hu, H. Lu, M. K. Mukerjee, M. Rabbi, et al., "NeuroPhone: brain-mobile phone interface using a wireless EEG headset," in Proceedings of the second ACM SIGCOMM workshop on Networking, systems, and applications on mobile handhelds, 2010, pp. 3-8.

[24] Z. A. Haq and Z. Hasan, "Eye-blink rate detection for fatigue determination," in 2016 1st India International Conference on Information Processing (IICIP), 2016, pp. 1-5.

[25] T. Ito, S. Mita, K. Kozuka, T. Nakano, and S. Yamamoto, "Driver blink measurement by the motion picture processing and its application to drowsiness detection," in Proceedings. The IEEE 5th International Conference on Intelligent Transportation Systems, 2002, pp. 168-173.

[26] J. Chen, W. Xuan, P. Zhao, U. Farooq, P. Ding, W. Yin, et al., "Triboelectric effect based instantaneous self-powered wireless sensing with self-determined identity," Nano energy, vol. 51, pp. 1-9, Jun. 2018.

[27] M. U. Anjum, A. Fida, I. Ahmad, and A. Iftikhar, "A broadband electromagnetic type energy harvester for smart sensor devices in biomedical applications," Sensors and Actuators A: Physical, vol. 277, pp. 52-59, Jul. 2018.

[28] J. C. Lin, "Biomedical applications of electromagnetic engineering," inEngineering Electromagnetics, ed: CRC Press, 2018, pp. 227-250.

[29] M. Serramito-Blanco, C. Carpena-Torres, J. Carballo, D. Piñero, M. Lipson, and G. Carracedo, "Anterior corneal curvature and aberration changes after scleral lens wear in keratoconus patients with and without ring segments," Eye & contact lens, vol. 45, pp. 141-148, Mar. 2019.

[30] A. Nakazawa, H. Kato, C. Nitschke, and T. Nishida, "Eye gaze tracking using corneal imaging and active illumination devices," Advanced Robotics, vol. 31, pp. 413-427, Jan. 2017.

[31] G. Salim and M. Zawawi, "Knee Joint Movement Monitoring Device based on Optical Fiber Bending Sensor," Journal of Telecommunication, Electronic and Computer Engineering (JTEC), vol. 10, pp. 25-29, Nov. 2018.

[32] R. K. Wangsness and R. K. Wangsness, Electromagnetic fields vol. 2: Wiley New York, 1979.

[33] D. Fleisch, A student's guide to Maxwell's equations: Cambridge University Press, 2008.

[34] H. A. Haus and J. R. Melcher, Electromagnetic fields and energy vol. 107: Prentice Hall Englewood Cliffs, NJ, 1989.

[35] C. Dodd and W. Deeds, "Analytical solutions to eddy‐current probe‐ coil problems," Journal of applied physics, vol. 39, pp. 2829-2838, May. 1968.

novel wearable sensors for biomechanical … · web viewnovel wearable sensors for biomechanical...

Documents