Sensors and Actuators A 148 (2008) 10–15

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

journa l homepage: www.e lsev ier .com/ locate /sna

Parametric design of yarn-based piezoresistive sensors for smart textiles

Ching-Tang Huanga,b, Chien-Fa Tangb, Ming-Chen Leeb, Shuo-Hung Changa,c,∗

a National Taiwan University, Department of Mechanical Engineering, No. 1, Sec. 4, Roosevelt Road, Taipei, Taiwanb Department of Products, TTRI, Taipei, Taiwanc Mechanical Lab., ITRI, Hsin Chu, Taiwan

a r t i c l e i n f o

Article history:Received 17 March 2008Received in revised form 29 May 2008Accepted 17 June 2008Available online 5 July 2008

Keywords:Yarn-based sensorSmart textiles

a b s t r a c t

In our previous paper, the yarn-based sensors are shown capable of measuring the respiration signals. Incomparing with fabric-based sensors, the yarn-based sensor is more comfortable in wearing, and easilyfabricated by conventional textile process. In this paper, we aim at the analysis of the design parameters ofthe yarn-based sensors that were fabricated by double wrapping the polyester, elastic, and piezoresistivefibers. The performances of the yarn-based sensor under different compositions have been experimentallyevaluated. It is shown that the soft-core yarn sensors achieve high gain factors with low linearity. Thesensors consisting of high-density piezoresistive fibers achieve high linearity with low gain factor. It wasalso shown that the nonlinearity of the sensors that is mainly due to the irregular characteristics of the

Piezoresistive yarn structure exits at the low strain region. As at the high strain region, the high linearity of the yarn-based sensors is found regardless of material compositions. To further reduce irregular characteristics of

ase t

1

atw[dapiod

wocbscf

cs

orabbitm

afpbebOd

0d

the yarn structure, to increfibers are suggested.

. Introduction

Smart textiles with embedded sensors to monitor, actuate anddapt to environment change become more and more popular inhese years [1]. One advanced study was to integrate smart texturesith semiconductors to develop a wearable-computing technique

2–4]. The other study was to develop smart textiles which canetect environmental change, and thus they can be used to measurend monitor the physiological conditions of the wearer. Althoughromising results were shown in utilizing the smart textures, those

ssues of complicated manufacturing process, performance deteri-ration after washing or repeated folding, and difficulty aestheticesign still hinder the way of commercialization.

One type of fabric-based sensors using piezoresistive fabricsas developed in which the fabrics were coated by a thin layerf piezoresistive materials, such as polypyrrole (PPy, a

∏-electron

onjugated conducting polymer) or a mixture of rubbers and car-

ons, on conventional fabrics [5–9]. The sensors behave like flexibletrain gauges which measured strain or stress and were used toapture posture or motion [6,10–12]. It was shown that the coatedabrics were highly dependent upon knitting or weaving topol-

∗ Corresponding author. Tel.: +886 2 23633863.E-mail addresses: cthuang.0528@ttri.org.tw (C.-T. Huang),

ftang.0277@ttri.org.tw (C.-F. Tang), mclee.d081@ttri.org.tw (M.-C. Lee),huochang@itri.org.tw, shchang@ntu.edu.tw (S.-H. Chang).

gssypi

ppp

924-4247/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.sna.2008.06.029

he preload and to choose the stiffer core yarn and the harder piezoresistive

© 2008 Elsevier B.V. All rights reserved.

gy and the performance of the sensors might degrade a lot afterepeated folding or washing. To improve the stability of the sensors,nother method was to knit conductive fibers with nonconductivease fibers [8,13,14]. The integration of conductive fibers with thease fibers can improve the sensing performance even after wash-

ng. However, only the knitting process can be adopted to fabricatehe sensors which might constrain the freedom when designing

odern clothes.Besides fabrics-based sensors, yarn-based sensors were another

pproach which could have the advantages such as more com-ortable in wearing, easier to be fabricated by conventional textilerocess, better in style design, higher space resolutions and possi-le to measure distributed strain. Instead of using fabrics as baselements, the yarn itself is the base element. The yarn is fabricatedy using piezoresistive fibers, elastic, and polyester fibers [15,16].ur previous paper showed that the yarn-based sensor using theouble wrapping method can achieve higher linearity than the sin-le wrapping approach. The sensor could measure the respiratoryignals precisely [16]. To extend the applicability of the sensor, moretudies should be performed to provide guideline in designing thearn-based sensors. The issues of using different core yarns andiezoresistive fibers are explored to provide more understandings

n designing yarn-based sensors.In this paper, the yarn-based sensors were fabricated by

olyester, elastic and piezoresistive fibers using the double wrap-ing methods with different compositions. Different polyester,iezoresistive fibers, and the fabrication process (twist per meter,

nd Actuators A 148 (2008) 10–15 11

TdmettItmtad

2

ppwiseoaddttiAwb

ytpratRt

cc

fdt

DmIp

3

icblwtmws

TN

T

TND

T

TND

T

TND

C.-T. Huang et al. / Sensors a

PM) are chosen as the design parameters. Performances of theesigned yarn-based sensors were experimentally evaluated byeasuring their resistance change under various elongation for lin-

arity and gain factor. To interpret the nonlinear characteristics ofhe sensor, an analysis method is proposed to calculate the force dis-ribution between the core yarn and the carbon-coated fiber (CCF).t was shown that the soft-core yarn sensors achieve high gain fac-ors at low linearity. The nonlinearity of the yarn-based sensors

ainly comes from the irregular characteristics of the yarn struc-ure specifically at the low strain region. Physical interpretationsre given to illustrate the behavior of the yarn-based sensor withifferent compositions.

. Fabrication of the yarn-based sensors

In this paper, the yarns were fabricated by wrapping theolyester fibers, elastic, and piezoresistive fibers into a skein. Theiezoresistive fiber was chosen as the CCF. The fabrication processas to combine polyester fibers with elastic fibers into a compos-

te core yarn. The length of the elastic fiber (Lycra, 22dtex) wastretched three times in the process before the combination. Thelastic fiber is to improve the elasticity of the sensing yarn. Thenne CCF was used to wrap the core yarn in the clockwise directionnd the other CCF is to wrap the core yarn in the counterclockwiseirection to form a double wrapping yarn as shown in Fig. 1. Theiameter of the yarn was less than 0.2 mm. The twist density ofhe CCF fiber warping on the core yarn is normally described byhe TPM. For example, TPM 450 means the length of every twists 1000/450 = 2.22 mm. The lower TPM implies the longer twist.lthough the single wrapping process can also be used, the doublerapping method was proven to be more effective as the yarn-

ased sensor [16].One objective of this paper is to identify better design of the

arn-based sensors. Since the yarn consists of the polyester, elas-ic, and piezoresistive fibers, two different kinds of commercialolyester fibers, 56dtex/144f and 56dtex/48f, were chosen to fab-icate the core yarn. Here the dtex denotes the gram per 10,000 mnd f denotes that number of filaments is contained. Moreover, twoypes of the CCF fibers (RESISTAT F901, MERGE S022 (24dtex) andESISTAT F901, MERGE D044 (49dtex)) were utilized to investigatehe piezoresistive effects on yarn-based sensors.

Besides using different types of the fibers, the TPM was alsohosen as another factor. The effects of the TPM under different

ombinations of polyester/piezoresistive fibers were investigated.

As shown in Table 1, nine different types of samples wereabricated. The samples are named as DXXX/YYf/ZZ, where XXXenotes the values of the TPM, YYf as the number of filaments ofhe polyester, and ZZ denotes the dtex of the CCF. For example,

sayms

able 1ine types of samples for experiments

ype D150/48f/24

PM (twist per meter) 150o. of filaments of polyester 48tex of the CCF (gram per 10 km) 24

ype D150/144f/24

PM (twist per meter) 150o. of filaments of polyester 144tex of the CCF (gram per 10 km) 24

ype D150/144f/49

PM (twist per meter) 150o. of filaments of polyester 144tex of the CCF (gram per 10 km) 49

Fig. 1. The structure of yarn-based sensors: (a) core yarn and (b) CCF.

150/48f/24 means that the TPM is 150, the number of the fila-ent of the polyester fibers is 48 and the dtex of the CCF fiber is 24.

n the next section, experiments were conducted to evaluate theerformances of the samples.

. Analysis method and experimental results

To evaluate the performance of the yarn-based sensors, exper-ments on resistance change under different elongation wereonducted in this section. The length of the samples is chosen toe 60 mm and the preload of 0.02 kg were used. The total stretched

ength was 14 mm. During the elongation, the force and resistanceere simultaneously measured. Tensile forces were recorded by

he Mini44 INSTRON and resistances were recorded by the Fluke189ulti-meter. After the measurement, the resistance and force curveere fitted by either the first or the second regression models as

hown in Tables 2 and 3. The first coefficient of the first regres-

ion model and its R2 value are used to evaluate the gain factornd linearity, respectively. The gain factor is the sensitivity of thearn-based sensor under the given strain. The higher the gain factoreans that the designed sensor can provide more sensitivity and

tronger noise resistance. Therefore, the gain factor and linearity

D275/48f/24 D450/48f/24

275 45048 4824 24

D275/144f/24 D450/144f/24

275 450144 144

24 24

D275/144f/49 D450/144f/49

275 450144 14449 49

12 C.-T. Huang et al. / Sensors and Actuators A 148 (2008) 10–15

Table 2The regression equations and R2 values of the CCFs

Resistance (M�) versus strain(�L/L) R2 Force (kg) versus strain(�L/L) R2

The first-order regressionCCF24 y = 18.601x + 1.998 0.983 z = 0.2843x + 0.0338 0.9345CCF49 y = 6.5659x + 1.026 0.9979 z = 0.6302x + 0.0305 0.9843

The second-order regressionCCF24 y = 11.819x2 + 15.844x + 2.097 0.985 z = −1.125x2 + 0.547x + 0.0240 0.9870CCF49 y = 4.544x2 + 5.506x + 1.064 0.9999 z = −1.1017x2 + 0.8872x + 0.0212 0.9965

Table 3The regression equations and R2 values of the yarns

Samples M� versus �L/Lelongation: 0–14 mm

R2 GF M� versus �L/Lelongation: 6–14 mm

R2 GF M� versus �L/Lelongation: 9–14 mm

R2 GF

D150/48f/24 y = 5.7271x + 0.5550 0.9878 5.7271 y = 6.4592x + 0.4325 0.9991 6.4592 y = 6.5305x + 0.4188 0.9983 6.5305D275/48f/24 y = 4.9440x + 0.5855 0.9917 4.9440 y = 5.4231x + 0.5035 0.9993 5.4231 y = 5.5663x + 0.4755 0.9991 5.5663D450/48f/24 y = 4.1860x + 0.6415 0.9882 4.1860 y = 4.778x + 0.5396 0.9984 4.778 y = 4.8999x + 0.5163 0.9977 4.8999D150/144f/24 y = 6.823x + 0.7307 0.9702 6.823 y = 8.3752x + 0.4646 0.9958 8.3752 y = 8.9787x + 0.3478 0.9983 8.9787D275/144f/24 y = 6.4759x + 0.7758 0.9598 6.4759 y = 8.2431x + 0.4719 0.9947 8.2431 y = 8.9324x + 0.3385 0.9989 8.9324D + 0.528D + 0.38D + 0.44D + 0.38

wt

eudafawmcta0qs1tCeiWofia

f

ortdo

fiiopftlUip

2

450/144f/24 y = 5.7812x + 0.7822 0.9658 5.7812 y = 7.2514x150/144f/49 y = 3.2004x + 0.4395 0.9935 3.2004 y = 3.5091x275/144f/49 y = 3.1103x + 0.5265 0.9866 3.1103 y = 3.5729x450/144f/49 y = 3.1772x + 0.4525 0.9895 3.1772 y = 3.5852x

ere used as the performance indices to evaluate the behavior ofhe sensor.

Five samples for each type listed in Table 1 were prepared forxperiments. The averaged measured data of the five samples weresed for analyses. To simplify the expression, the strain (�L/L) wasenoted as x, the resistance denoted as y and the force denoteds z throughout the paper. Before investigating the effects of dif-erent compositions, the piezoresistive characteristics of the CCFlone were measured first. Two CCFs with the dtex of 24 and 49ere tested, denoted as the CCF24 and CCF49, respectively. Theeasured resistance versus strain is shown in Fig. 2. The statisti-

al data is listed in Table 2. For the resistance curve of the CCF24,he R2 values for the first and second regression models are 0.983nd 0.985, respectively. For that of the CCF49, the R2 values are.9979 and 0.9999, respectively. It shows that linear relation is ade-uate in describing the relationship between the resistance and thetrain for the CCFs. As recalled that 24dtex denotes the 24 g per0,000 m, the linear density of the CCF49 is about two times ofhat of the CCF24. The circumference of the cross-section for theCF49 is about

√2 times of that of the CCF24. The larger circumfer-

nce allows the larger cross-section area for current to flow, whichmplies the resistance of the CCF49 is smaller than that of the CCF24.

hen both CCFs are under the same strain, the resistance change

f the CCF49 is smaller which results in lower gain factor. From therst regression models listed in Table 2, the gain factors of 18.60nd 6.57 were found for the CCF24 and CCF49, respectively.

As indicated in Table 2 and Fig. 2(b), the relation between theorce and the strain for both CCFs are better represented by the sec-

Fig. 2. The measured values under different strain for CCF24 and C

5 0.9916 7.2514 y = 8.0497x + 0.3735 0.9975 8.049771 0.9997 3.5091 y = 3.5672x + 0.3759 0.9997 3.567274 0.9992 3.5729 y = 3.661x + 0.4304 0.9992 3.66129 0.9986 3.5852 y = 3.7252x + 0.3558 0.9992 3.7252

nd regression equation with the R2 value of 0.9870 and 0.9965,espectively. Since the linear density of the CCF49 is about twoimes of that of the CCF24 and the stiffness is proportional to linearensity, it is obvious that the CCF49 has a higher stiffness than thatf CCF24.

The rational of obtaining the regression model of the CCF alonerst before evaluating the performances of the yarn-based sensor

s that one can separate the force distribution of the CCF with thatf the core yarn using the regression model in Table 2. Since theiezoresistive properties of the yarn-based sensors come directlyrom the CCF, the resistance change of the sensor is directly relatedo the change of the CCF. From the resistance change, one can calcu-ate the strain and thus determine the force on the CCF from Table 2.nderstanding the force exerted on the CCF would be helpful to

llustrate the characteristics of the yarn-based sensor. The analysisrocedure is proposed as following.

1. Measure the resistance of the yarn-based sensor under dif-ferent elongations. The results are summarized in Table 3where the regression equations, R2 values and gain factors arelisted.

. Substitute the resistance of the measured data into the first-

order regression function listed in Table 2 and calculate the strainof the CCF. Since the CCFs in the yarn-based sensors could beseen as two parallel resistors, the measured resistance shouldbe multiplied by 2 to get the resistance for the single CCF beforesubstituted into the equations.

CF49: (a) resistance versus strain and (b) force versus strain.

C.-T. Huang et al. / Sensors and Actuators A 148 (2008) 10–15 13

F 50/48f

3

4

5

Ftmc

4

fcnm

4

otfepmtotsft

terlssfTmtbs

hdifDitl

4

tDtCCt

F

ig. 3. Estimated force and resistance change under different strain for samples D1

. Use the second-order regression model for the strain and forceshown in Table 2 to calculate the longitudinal force of the CCF.

. The force in (3) is multiplied by 2 to obtain the summing forceof two CCFs.

. The force on the core yarn can be calculated by subtracting theforce on the CCFs from the total applied force.

The results for the samples with TPM 150 were shown inigs. 3 and 4. These results will be discussed in the next sectiono illustrate the characteristics of the yarn-based sensor. Further-

ore, the linearity and gain factor for the samples with differentompositions and TPM will be analyzed.

. Analyses of the yarn-based sensors

In this section, the parameter analysis on the sensors with dif-erent core yarns and CCFs are discussed. The effects of differentore yarn, CCF, and TPM are discussed first. Then the nonlinear phe-omenon found in the sensors is explained based on the regressionodel in Table 3.

.1. The effects of core yarn

The samples with the same linear density but different kindsf polyester (56dtex/144f and 56dtex/48f) are used to investigatehe effect of the core yarn. Fig. 3 shows the experimental resultsor the samples D150/48f/24 and D150/144f/24. The only differ-nce between the two samples is the number of the filaments ofolyester fibers in the core yarn. The cross-section of each fila-ent for the 56dtex/144f is 1/3 of that of the 56dtex/48f since

he number of the filaments for 56dtex/144f is three times of that

f the 56dtex/48f. It is obvious that the yarn-based sensor usinghe 56dtex/144f could result in softer structure and can be easilytretched and compressed. It could be anticipated that the loadingorce distributed in the core yarn for the 56dtex/144f is smaller thanhat for the 56dtex/48f when the sensor is under the same strain.

tit

T

ig. 4. Estimated force and resistance change under different strain for samples D150/144

/24 and D150/144f/24: (a) force versus strain and (b) resistance versus strain.

The experimental results shown in Table 3 and Fig. 3 (b) implyhat the soft-core yarn (D150/144f/24) has higher GF but lower lin-arity. As shown in Fig. 3(a), the top two curves are the experimentalesults for the D150/48f/24 and D150/144f/24. It is found that theoading forces of the two samples are almost the same under theame strain. Because the 56dtex/144f of the core yarn can be easilytretched, the CCF in the sample D150/144f/24 should afford higherorce which is demonstrated by the lower two curves in Fig. 3 (a).he forces distributed in the CCFs are calculated by the analysisethod proposed in the previous section. Consequently, the resis-

ance change of the sample D150/144f/24 with 56dtex/144f shoulde higher under the same strain than that of the D150/48f/24 ashown in Fig. 3(b).

The gain factor 6.823 of the D150/144f/24 is found which isigher than that of the D150/48f/24, which is 5.7271 because theeformation of the CCF in the D150/144f/24 is larger. The linear-

ty (R2 values) can be observed from Table 3 where the R2 valueor the D150/144f/24 is 0.9702 which is smaller than that of the150/48f/24 (0.9878). The issues of the linearity will be discussed

n Section 4.4. Note that by utilizing the proposed analysis method,he difference of the gain factor between the two samples can beogically explained.

.2. The effects of CCFs

In this section, the effects of using different CCFs are inves-igated. Fig. 4 shows the experimental results for the samples,150/144f/24 and D150/144f/49. The only difference between the

wo samples is the line density of the CCF. Since the stiffness of theCF 49 is higher than that of the CCF 24, it can be anticipated that theCF in the sample D150/144f/49 would afford larger loading forcehan that in the D150/144f/24 under the same strain. This observa-

ion can be found in Fig. 4(a) where the stiffness of the D150/144f/49s higher than that of the D150/144f/24 and the force of the CCF inhe D150/144f/49 is higher, too.

The GFs for the samples with different CCFs are found inable 3. The GF of the D150/144f/49 (3.2004) is lower than that

f/49 and D150/144f/24: (a) force versus strain and (b) resistance versus strain.

1 nd Actuators A 148 (2008) 10–15

oit2lttc

cafittTclsmsrTi4

4

DrDrtalTcrish

CDmoC

4

FswtfitetreooI

Fa

attdc

aabcesrrcaztacp

5

dppbrs

4 C.-T. Huang et al. / Sensors a

f D150/144f/24 (6.823). Even though the force loading on the CCFn the D150/144f/49 is higher under the same strain, the gain fac-or of the CCF 49 alone (6.57) is much lower than that of the CCF4 alone (18.6) as shown in Table 2. The compensation of largeroading force on the CCF in D150/144f/49 makes the gain factor ofhe D150/144f/49 is about half of that of the D150/144f/24. Again,he observation can be illustrated after the force distribution isalculated by the analysis method.

The comparisons of the linearity between the two samplesan be found from Fig. 4 (b) and Table 3. The D150/144f/49 canchieve higher linearity with the R2 being equal to 0.9935. Oneactor is that the linearity of the CCF 49 alone is high as shownn Table 1. The other is the weaker effect of the lateral force. Ashe tensile stress is applied to the sensor as shown in Fig. 1, bothhe longitudinal force and the lateral force should be considered.he longitudinal force causes the elongation, but the lateral forceompresses the core yarn to change the shape. The effect of theateral force is the alleviation of linearity. Because the CCF49 istiffer, it implies the core yarn is more constrained after yarn for-ation and there is less space to be compressed. Therefore, the

hape change due to the lateral force is relative smaller and theelative nonlinear effects caused by the lateral force are smaller.hat is why the sensor with CCF49 can provide higher linear-ty. More discussion on the linearity will be provided in Section.4.

.3. The effects of the TPM

As shown in Table 3, the GFs for the samples D150/144f/24,275/144f/24, and D450/144f/24 are 6.823, 6.4759 and 5.7812,

espectively. Also, the GFs for the samples D150/48f/24,275/48f/24, and D450/48f/24 are 5.7271, 4.9440 and 4.1860,

espectively. Obviously, GF decreases as the TPM increases forhe samples with CCF24. The phenomenon can be explaineds follows. When the yarn-based sensor is stretched, both theongitudinal force and the lateral force are exerted on the yarns.he longitudinal force causes the elongation, but the lateral forceompresses the core yarn to change the shape. The shape changeeduces the deformation of the CCFs. Since the larger TPM resultsn higher lateral forces and thus the shape change under the sametrain is larger. Therefore, the gain factors of the samples withigher TPM are smaller.

The effect of TPM could not be found in the samples using theCF49. In Table 3, the GFs for the D150/144f/49, D275/144f/49 and450/144f/49 are 3.2004, 3.1103 and 3.1772, respectively. Becauseost of the loading force is afforded by the CCF49, the shape change

f the core yarn is not so significant to affect the deformation of theCF 49.

.4. Nonlinearity of the yarn-based sensor

The nonlinear property between the resistance and the strain inigs. 3 and 4 should be investigated to enhance the feasibility of theensor implementation. As indicated in Section 2, an elastic fiberas combined with the polyester fiber into the core yarn to provide

he elasticity for the sensors. The side effect of the adding elasticber is that the sensor is crimped without the preload. Even withhe preload, it cannot be ensured that the CCF is fully stretchedspecially when the strain is small. Fig. 5 shows the pictures ofhe sensor under the elongation equals to 1 mm, 3 mm, and 14 mm,

espectively. As shown in Fig. 5(a), the CCF is still crimped when thelongation equals to 1 mm. Under this condition, the force exertingn the CCF is comparatively small and most of the force is appliedn the core yarn. Therefore, the resulted resistance change is little.t explains why most of the curves of the resistance versus strain

Rti

t

ig. 5. The pictures of the yarn-based sensors at elongation of 1 mm (a), 3 mm (b),nd 14 mm (c).

re flatter at the low strain region in Figs. 3 and 4. The region wherehe CCF is not fully stretched is called the irregular zones sincehe deformation of the sensor is not very regular. Depending uponifferent composition in the yarn-based sensor, the irregular zonean be quite different.

As the elongation becomes larger, the CCF is no more crimpednd is fully stretched as shown in Fig. 5(b) and (c). It could also benticipated that the linearity is improved when the sensor is testedeyond the irregular zone. The concept can be further proved byalculating the linear regression equations and R2 values for thelongation from 6 to 14 mm and from 9 to 14 mm, where the corre-ponding strain varies from 8.3% to 23.3% and from 13.3% to 23.3%,espectively. As shown in Table 3, the R2 values of the first-orderegression equations were all higher than 0.99 regardless of theompositions of the sensors. It demonstrates that the sensor canchieve better performance if it is operating out of the irregularone. It is suggested that the preload can be increased in ordero avoid the irregular zone and thus the higher linearity can bechieved. Moreover, the irregular zone can be reduced when theore yarn is stiffer and the CCF is harder. These observations canrovide some guidelines in designing the yarn-based sensors.

. Conclusions

The parameter analyses of the yarn-based sensor are con-ucted in this paper. The yarn-based sensors were fabricated byolyester, elastic and piezoresistive fibers using the double wrap-ing methods. Nine different samples were fabricated and testedy measuring the resistance change under variable elongation. Theesistance curve was fitted by either the first or the second regres-ion model. The first coefficient of the first regression model and its

2 values are used to evaluate the gain factor and linearity, respec-ively. The linearity and gain factor were used as the performancendices to evaluate the behavior of the sensor.

Because the yarn-based sensor consists of the core-yarn andhe piezoresistive CCF fibers, the behavior of the sensor is hard

nd Act

tcWToidpttrifpaf

R

[

[

[

[

[

[

[

B

CtNSRiimm

CMis

MCnb

SvoCstsCN(

C.-T. Huang et al. / Sensors a

o explain. An analysis method is proposed in this paper to cal-ulate the force distribution between the core yarn and the CCF.

ith the proposed method, the effects of the core yarn, CCF andPM are discussed. It is demonstrated that the sensor consistingf soft-core yarns can achieve high gain factors but low linear-ty. The sensor consisting of piezoresistive fibers with high linearensity can achieve high linearity but the gain factor is low. Thehenomenon is explained by the analysis method. It is also shownhat the nonlinearity of the yarn-based sensors mainly comes fromhe irregular characteristics of the yarn structure at the low-strainegion. As the sensor operates at the high-strain region, the linear-ty of the yarn-based sensors is improved even the sensors wereabricated with different compositions. Some design guidelines arerovided to reduce the irregular zone. One is to increase the preloadnd the other is to choose the stiffer core yarn and harder CCF inabricating the yarn-based sensor.

eferences

[1] J. Farringdon, A.J. Moore, N. Tilbury, J. Church, P.D. Biemond, Wearable sen-sor badge and sensor jacket for context awareness, in: IEEE ISWC’99, PhilipsResearch Laboratories, UK, 1999, pp. 107–113.

[2] R.B. Katragadda, Y. Xu, A novel intelligent textile technology based on siliconflexible skins, electrical and computer engineering, in: ISWC, Wayne State Uni-versity, Detroit, MI, USA, 2005, pp. 78–81.

[3] J.B. Lee, V. Subramanian, Weave patterned organic transistors on fiber for E-textiles, IEEE Trans. Electron. Devices 62 (2005) 269.

[4] A. Bonfiglio, D. De Rossi, T. Kirstein, I. Locher, F. Mameli, R. Paradiso, G. Vozzi,A feasibility study of yarns and fibers with annexed electronic functions: theARIANNE project, in: A. Lymberis, D. de Rossi (Eds.), Wearable e-Health Systemsfor Personalized Health Management, IOS Press, Amsterdam, The Netherlands,2004, pp. 324–329.

[5] D. De Rossi, F. Carpi, F. Lorussi, A. Mazzoldi, E.P. Scilingo, A. Tognetti, Electroac-tive fabrics for distributed, conformable and interactive systems, IEEE Sens.(2002).

[6] F. Lorussi, E.P. Scilingo, M. Tesconi, A. Tognetti, D. De Rossi, Wearable sensing gar-ment for posture detection, rehabilitation and tele-medicine, in: Proceedingsof the Fourth International IEEE EMBS Special Topic Conference on InformationTechnology Applications in Biomedicine, Birmingham, UK, April 24–26, 2003,pp. 28–290.

[7] R. Wijesiriwardana, T. Dias, S. Mukhopadhyay, K. Mitcham, Knitted straingauges, in: The SPIE International Symposium on Microtechnologies for theNew Millennium Maspalomas, Canary Islands, Spain, May 19, 2003, pp.5116–5131.

[8] R. Wijesiriwardana, K. Mitcham, T. Dias, Fibre-meshed transducers based realtime wearable physiological information monitoring system, in: Proceedings ofthe eighth International Symposium on Wearable Computers (ISWC’04), 2004,

pp. 40–47.

[9] D. De Rossi, F. Carpi, F. Lorussi, R. Paradiso, E.P. Scilingo, A. Tognetti, Electroactivefabrics and wearable man-machine interfaces, in: Wearable Electronics andPhotonics, Woodhead Publishing Limited, UK, 2005.

10] E.P. Scilingo, F. Lorussi, A. Mazzoldi, D. De Rossi, Strain-sensing fabrics for wear-able kinaesthetic-like systems, Sens. J. IEEE 3 (4) (2003) 460–467.

ecoMoP

uators A 148 (2008) 10–15 15

11] D. De Rossi, F. Lorussi, A. Mazzoldi, P. Orsini, E.P. Scilingo, Monitor body kine-matics and gesture through sensing fabrics, in: Proceedings of the First AnnualInternational Conference On Micro-technologies in Medicine and Biology, Cen-tro Interdipartmentale di Ricerca, Pisa University, 2000, pp. 587–592.

12] F. Lorussi, W. Rocchia, E.P. Scilingo, A. Tognetti, D. De Rossi, Wearable, redundantfabric-based sensor arrays for reconstruction of body segment posture, IEEESens. J. 4 (6) (2004) 807–818.

13] R. Wijesiriwardana, T. Dias, S. Mukhopadhyay, Resistive fiber-meshed trans-ducers, in: Proceedings of the IEEE International Symposium on WearableComputing, 2003, pp. 200–205.

14] R. Wijesiriwardana, Inductive fiber-meshed strain and displacement transduc-ers for respiratory measuring systems and motion capturing systems, IEEE Sens.J. 6 (3) (2006) 571–579.

15] C.-T. Huang, C.-F. Tang, C.-L. Shen, A wearable textile for monitoring respiration,using a yarn-based sensor, in: ISWC’06, 2006, pp. 141–142.

16] C.-T. Huang, C.-F. Tang, C.-L. Shen, S.-H. Chang, A wearable yarn-based piezo-resistive sensor, Sens. Actuators A: Phys. (2007).

iographies

hing-Tang Huang graduated from National Taiwan University in 1988, and receivedhe M.S. degree in mechanical engineering from National Taiwan University in 1993.ow, he is a Ph.D. student in mechanical engineering, National Taiwan University.imultaneously, he is also the chief of system integration section, Taiwan Textileesearch Institute (TTRI). The major research topic is about the smart textile, focus-

ng on how to develop sensors and actuators, which are tiny and could be embeddedn the textile, or the textiles are sensors or actuators themselves, and how to imple-

ent the smart clothing with miscellaneous sensors and actuators on the basis ofaintaining the traditional comfort.

hien-Fa Tang received the MSc. degree in textiles from the UMIST (University ofanchester Institute of Science and Technology), UK, in 1997. He has been working

n the TTRI since 1990. His research interests include electrical interface, biomedicalignal processing, digital/analog circuit design, and system control.

ing-Chen Lee graduated from Electronics College Department, DE Lin Juniorollege in 2003, and will receive the B.S. degree in Department of Electronic Engi-eering, DE Lin Institute of Technology in 2008. He majors in electronics and haseen an assistance engineer in TTRI since 2001.

huo-Hung Chang received the BS degree in 1974 from National Chen Kung Uni-ersity, Taiwan and the MS and PhD degrees in 1981 and 1985 from the Universityf Cincinnati, USA. From 1984 to 1990 he worked at the IBM T.J. Watson Researchenter in Yorktown Heights, NY, involved in advanced computer peripheral devices,uch as printing, data storage and information display. Since 1990, he has been athe Department of Mechanical Engineering, National Taiwan University (NTU). Heerved as the Director of the Nano-Electro-Mechanical Systems (NEMS) Researchenter and the Deputy Director of the Center for Nano-Science and Technology inTU. He was a visiting scholar with National Institute of Standards and Technology

NIST) and the Stanford University in 2000 and 2003, respectively. His research inter-

sts involve electro-elasticity theory, sensors and actuators, nanometer positioning,arbon nanotubes, carbon nanocoils and nano-mechanics. Dr. Chang is a memberf the IEEE, American Society of Mechanical Engineers (ASME), Chinese Society ofechanical Engineers (CSME), Chinese Society of Mechanism and Machine The-

ry, and Chair of the Member China-Taipei, The International Federation for theromotion of Mechanism and Machine Science (IFToMM).