Watch my Painting: The Back of theHand as a Drawing Space forSmartwatches

Maximilian SchrapelHuman-Computer InteractionGroupLeibniz University HannoverHannover, Germanyschrapel@hci.uni-hannover.de

Steffen RyllHuman-Computer InteractionGroupLeibniz University HannoverHannover, Germany

Florian HerzogHuman-Computer InteractionGroupLeibniz University HannoverHannover, Germany

Michael RohsHuman-Computer InteractionGroupLeibniz University HannoverHannover, Germanyrohs@hci.uni-hannover.de

Permission to make digital or hard copies of part or all of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies bear this notice and the full citationon the first page. Copyrights for third-party components of this work must be honored.For all other uses, contact the owner/author(s).CHI ’20 Extended Abstracts, April 25–30, 2020, Honolulu, HI, USA.© 2020 Copyright is held by the author/owner(s).ACM ISBN 978-1-4503-6819-3/20/04.DOI: https://doi.org/10.1145/3334480.3383040

AbstractSmartwatches can be used independently from smart-phones, but input tasks like messaging are cumbersomedue to the small display size. Parts of the display are hid-den during interaction, which can lead to incorrect input.For simplicity, instead of general text input a small set ofanswer options are often provided, but these are limitedand impersonal. In contrast, free-form drawings can an-swer messages in a very personal way, but are difficult toproduce on small displays. To enable precise drawing in-put on smartwatches we present a magnetic stylus that istracked on the back of the hand. In an evaluation of sev-eral algorithms we show that 3D position estimation witha 7.5×20 mm magnet reaches a worst-case 6 % relativeposition error on the back of the hand. Furthermore, theresults of a user study are presented, which show that inthe case of drawing applications the presented technique isfaster and more precise than direct finger input.

Author KeywordsDigital Pens, Mobile Interaction, Around-Device Interaction

CCS Concepts•Human-centered computing→ Interaction devices;Ubiquitous and mobile devices;

Introduction

Figure 1: Intended application: Apicture is drawn with the Device onthe back of the hand.

Figure 2: Comparison of theprototype (center) versus normalpens. The dimensions of aconventional pen provide a naturalwriting experience.

Modern smartwatches are mainly used in sporting activi-ties, health tracking, and with simple interactions such asanswering phone calls, navigation, or music control. Dueto their functionality of phone-independent mobile commu-nication they can be used in a variety of applications, butface challenges associated with the small interaction areaof the touch display. For example, tasks like text messagingare still difficult, because the finger covers several keys andusers cannot see what they are actually typing [18].

To address this challenge, many researchers proposednovel interaction designs based on software and hardwareprototypes to improve the usability of ultra small touch-screens. As an alternative option to typing, drawing is avery personal way to reply to incoming messages or to cre-ate, for example, an individualized digital birthday card [16].However, drawing requires very precise input, which ampli-fies the challenges of small displays, e.g., with regard to thefat finger problem [35].

Interactions in the spatial vicinity of a display are particu-larly suitable to solve occlusion problems on tiny devices.To this end we developed a pen-based input device at thesize of a conventional pen that allows cursor-like input onthe back of the hand. The pen contains a capacitive pen-tipsensor to detect touch events on the hand and a permanentmagnet to allow the smartwatch magnetometer to track theposition of the pen.

This paper evaluates tracking algorithms and derives opti-mal parameters to enable precise paintings on smartwatchdisplays. In addition, the results of an initial user study arepresented to prove the concept and to evaluate the per-formance compared to direct finger input on smartwatchdisplays.

Related WorkMethods for counteracting the fat finger problem on tinydisplays can be divided into software and hardware ap-proaches. By extending the software of the device, sim-ple interaction techniques can be realized that do not re-quire further equipment. To name a few, by enlarging ar-eas of interest, precise input can be generated [18, 23, 26].Also, splitting the keyboard into smaller parts has been ex-plored [7, 8, 14, 27], as well as tap and gesture typing [12],cursor entry [31], or one-handed input via moving the wrist [10].

In order to avoid occluding the display, additional hardwarecan extend the interaction space. There are approachesthat add hardware to the finger segments [3, 6, 24, 13, 28,34, 36, 38], sense the wristband or smartwatch bezel forinput [11, 21, 25, 29, 33, 40], or use a movable mechan-ical display for simple interactions like pan, twist, tilt, andclick [39]. Other proposals add touch sensitive wire intoclothing [29], use the skin [22] and the back of the hand forinput [4, 17], or enable mid-air interactions around the de-vice [15, 20, 32]. Several projects use magnetic field sens-ing [3, 6, 15, 20, 24, 28, 34], which has also been appliedon stylus pens [1, 41]. There has also been research onfinding optimized algorithms for permanent magnets withdifferent shapes for tracking in 2D [5] and using magne-tometer arrays around the device for 3D position estima-tion [9].

However, these implementations do not optimize trackingfor precise interactions, like drawing on the back of thehand. For this purpose, Wu et al. combine a monocularcamera and a 6-DoF motion tracking sensor to achieve veryprecise drawings [37]. Without position tracking, Schrapelet al. have shown that high recognition rates on handwrit-ten digits can be achieved with stylus pens [30]. Both tech-

niques use paper as the interaction area and do not con-sider the back of the hand for input.

While previous work focused on general interactions, weshow how the back of the hand can be used to enable pre-cise cursor-like input and demonstrate how the interactioncan be optimized for painting with digital pens.

Position Tracking

Figure 3: Study setup and testhousing for the magnets. TheHuawei Watch was placed in thecoordinate system so that themagnetometer is at its origin.

6%

5%

4%

3%

2%

1%

0%

Relative position error

Figure 4: Visualization of therelative error distribution on theback of the hand with a 7.5x20 mmmagnet and 3D magnet positiontracking: Near the magnetometerthe error increases due to sensorsaturation. The relative error iscalculated by dividing the absoluteerror of the calculated distance bythe real distance.

Since painting requires a particularly high tracking preci-sion, we evaluate known algorithms for magnetic tracking.

Algorithm SelectionTo identify the best tracking algorithm, three different ap-proaches from related work [1, 5, 9, 20, 41] were imple-mented on a Huawei Watch 2. In the selection process forthe most suitable tracking algorithm, the intended use witha stylus pen was especially considered. The three candi-date algorithms we considered were the ones by Abe etal. [1], by Camacho and Sosa [5], and by Yoon et al. [41].

We first chose the 2D tracking algorithm by Abe et al. [1],assuming that the back of the hand approximates a flatplane. When all influences of terrestrial magnetism areconstant, the magnetic moment can be eliminated usingCoulomb’s law, and a root-finding problem remains. Incontrast to Abe et al., who apply the bisection method, weuse Newton’s method on all algorithms, as it requires lesscomputational power, which is advantageous for mobile de-vices. While Abe et al. only consider the height of the mag-net [1] Camacho and Sosa also use its shape in their equa-tions [5]. Since cylindrical shapes can easily be embeddedin stylus pens, this shape and approach was selected.

Furthermore, we investigated the 3D tracking algorithmby Yoon et al. [41] which was used in various projects [9,20]. In contrast to the other chosen methods, this approach

takes into account the magnetic moment, which determinesthe orientation of the magnet. However, the orientation ofthe magnet in three dimensions is required, which Yoon etal. calculate using an additional magnetometer [41] andFan et al. [9] as well as McIntosh et al. [20] by placing sev-eral magnetometers around the interaction surface.

Algorithm EvaluationTo examine the optimal magnetic parameters and to eval-uate the performance of the selected algorithms with themagnetometer embedded in the watch, a preliminary exper-iment was conducted. Based on the dimensions chosen byAbe et al. [1], cylindrical magnets with a diameter of 8 mmand heights of 16, 20, and 32 mm with a fixed magnetiza-tion degree of N48 were selected. Since a higher magne-tization results in a higher magnetic flux density, a magnetwith a diameter of 7.5 mm and a length of 20 mm with amagnetization of N52 was also used in the experiments.

The first experiment was intended to measure the accu-racy of the algorithms in 2D space. The Huawei Watch 2was placed in a planar coordinate system with the embed-ded 9-DOF smartwatch IMU in the origin. Within the pro-posed interaction area the cylindrical permanent magnet(diameter 8 mm, height 20 mm) was moved in 1 cm stepswithin a grid of 4 to 12 cm in x-direction and -5 to 1 cm iny-direction. The grid size was selected according to the av-erage of anthropometric measurements of the back of thehand [2]. At each position, with the magnet’s north pole to-wards the ground, the calculated distance was logged for allthree algorithms. Deviations, resulting from inaccurate mag-net orientation tracking for the 3D tracking algorithm wereeliminated due to the chosen test setting.

A second experiment was conducted to investigate the op-timal magnetic parameters. The previous test was repeatedfor different magnets with the 3D tracking algorithm. Based

on the results, the two most promising magnets were eval-uated again in an tilt angle of 45°, which corresponds to amore realistic usage for stylus pens. Based on the final se-lection, the influence of the magnet’s offset in z-directionwas then investigated by measuring the positions again withan offset of 5, 10, 15, and 20 mm on the z-axis.

Tracking Results

7.5x20 N52 8x20 N48Algorithm

0

5

10

15

Rel

ativ

e Err

or [

%]

Algorithm Accuracy with 45° inclination

xmax=4cm

7.5x20 N520mm 5mm 10mm 15mm 20mm

Magnet distance to ground

0

2

4

6

8

10

Rel

ativ

eErr

or

[%]

Relative error vs. magnet z-offset

7.5x20 N52 8x16 N48 8x20 N48 8x32 N48Magnet dimension

0

5

10

15

20

25

30

Rel

ativ

eErr

or

[%]

Magnet Comparsion

xmax=5cm

Abe et al. [1] Camacho et al.[5] Yoon et al.[41]Algorithm

0

10

20

30

40

50

Rel

ativ

eErr

or

[%]

Algorithm Accuracy

b

a

d

c

Figure 5: Position trackingevaluation: 3D tracking with a7.5×20 mm N52 magnet achievesthe best performance.

The general algorithm accuracy comparison Figure 5ashows that 3D tracking clearly outperforms 2D approacheseven on a planar surface. Likewise, the back of the handdoes not offer a completely flat surface, making 3D positiondetermination particularly suitable for the intended applica-tion. We observed that the 2D algorithm by Camacho andSosa [5] achieves more stable error rates than Abe et al. byincorporating the shape of the magnet into the calculations.However, a post-hoc test with Bonferroni correction showedno significant differences between the 2D algorithms whileboth differ significantly from 3D tracking.

Figure 5b shows that the strength of the magnetic fieldhas a major impact on accuracy. If the magnetic field istoo weak, the algorithm cannot find an optimal solution atevery position within the interaction area. As a result themagnet [8×16 mm; N48] can only be used within a radiusof 5 cm on the proposed interaction area. Conversely, ifthe magnetic field is too strong (see [8×32 mm; N48]), themagnetometer becomes saturated, which leads to inaccu-rate results. Furthermore, the watch itself has a magneticmounting for the charging cable that influences the embed-ded sensor. Although this constant factor can be eliminatedfrom the measurements, it still has an influence on satura-tion and noise that results in less accurate position tracking.A post-hoc test with Bonferroni correction could also showa significant difference between the magnet with a size of8x32 mm and the other magnets.

For the two remaining magnets, Figure 5c shows that aninclination also affects the positioning accuracy. It can beassumed that the magnet of the smartwatch mounting herealso influences the tracking algorithm. Due to the slightlyhigher flux density of the magnet [7.5×20 mm; N52] accu-rate positions can still be calculated, while the other magnet[8×20 mm; N48] is limited to a radius of 4 cm.

Figure 5d shows the influence of the z-position on the com-puted 2D position. A relatively stable accuracy is achievedover the full range, whereby an offset starting from 10 mmdiffers significantly compared to the initial position, accord-ing to a post-hoc test with Bonferroni correction.

This is due to the fact that the magnetic flux density de-creases with increasing distance. Therefore it is importantto consider the outliers, which up to 15 mm achieve a max-imum relative error of 6 % within the proposed interactionarea. Figure 4 visualizes the tracking performance overthe intended interaction area. Since the absolute error in-creases with a rising distance, direct mapping of the posi-tion data results in inaccurate positioning of the cursor onthe watch. Therefore we use indirect mapping.

PrototypeFrom the previous evaluation we conclude that a cylindri-cal magnet [7.5×20 mm; N52] is able to track the positionover the back of the hand with a maximum relative posi-tion error of 6 %. The found dimensions fit into a conven-tional pen, which is important to achieve a natural writingexperience. In addition to the small form factor, accurateinteraction is also required. For this purpose, a touch sen-sor (AT42QT1010) on the PCB is connected to the pen tip,which communicates its state to the smartwatch via Blue-tooth Low Energy (version 5) upon contact with the skin.The round shape of the pen tip and the gold plating ensure

a comfortable touch experience and an optimal contact withthe skin even in an inclined position. Figure 6 shows theindividual components.

A 6-DOF IMU (BMI160) is integrated to measure the orien-tation of the stylus, which requires an initial calibration onfirst use. For this purpose the prototype must be held per-pendicular to the back of the hand for one second. Basedon the sensor data measured at 100 Hz, the position is thencalibrated with Madgwick’s AHRS [19]. A magnetometer(BMM150) was also installed for test purposes, which can-not be used for orientation estimations due to the strongmagnetic field. This leads to an accumulated drift over time,which, however, does not noticeably influence the accuracydue to the low-noise sensor data over the short durationof an interaction [41]. To alleviate such tracking drifts, anindirect mapping is applied which does not require furthercalibration. In addition, it is thereby not necessary to furtheradjust the tracking area to the size of the hand.

IMU

BLE

5.0

BLE

5.0

Touch

Pen t

ipBLE

5.0

Bat

tery

Mag

net

Figure 6: Prototype in detail: The[8x30] mm PCB, the gold-platedpen tip and the side view of thepen.

Finger

Stylus

Easy Medium Hard

Figure 7: Drawing test results ofone participant.

The 7.5×20 mm magnet is mounted at a distance of 15 mmfrom the pen tip, also serving as a pen handle that restrictsthe degrees of freedom in which the pen is held. This isadvantageous as the corresponding yaw angle of the pro-totype can only be calculated using a magnetometer. In ad-dition, the pen’s center of gravity is shifted further towardsthe tip, which positively affects the writing comfort. In totalthe prototype has a weight of 18 g, of which 12 g are usedfor the 180 mAh battery. It can be continuously operated for12 hours without recharging the battery.

For demonstration purposes an example painting applica-tion was implemented on the Huawei Watch 2. A cursor onthe display shows the current position. As soon as the pentip touches the back of the hand, a line is drawn.

User StudyA small user study was conducted to compare the pen-based interaction on the back of the hand with direct fingerinput for target acquisition and drawing interactions on tinydisplays. Twelve volunteers aged 18 to 52 years (mean agex = 29.0, σ = 11.7), including 4 women and 8 men wererecruited. All participants were right-handed and 6 of themowned a smartwatch.

After an initial questionnaire with demographic questionsa short introduction was given by the demonstrator. Theparticipants could then test the application for as long asthey wanted to get familiar with the interaction technique.

Then a pointing experiment was performed to test howprecise and fast users can select randomly displayed tar-gets with a radius r=10 px (1 mm). After ten trials with thefinger and ten trials with the pen, each participant had toselect ten names from vertically aligned random lists witheach technique. The objective was to compare list scrollingby drawing with the pen on the back of the hand againstscrolling on the display with the index finger. Together, bothexperiments provide an insight into how quickly and pre-cisely users can interact with each technique.

A final experiment was intended to find out how accuratelyboth methods can be used to create drawings. For this pur-pose, the participants had to trace various shapes as ac-curately as possible. A periodic rectangle function was dis-played in three levels of difficulty, i. e., with three, four andfive repetitions, as shown in Figure 7. Additionally, two let-ters (M, ß) had to be drawn. All variants of drawings werepresented in random order and measurements of time andposition were taken from the beginning of the drawing. Thewhole study took about 45 minutes on average.

ResultsThe results of the user study are visualized in Figure 8. Theupper graph (a) shows the results of the pointing experi-ment. While the finger can be used to interact much faster,a higher error rate occurs than with the stylus. This be-haviour also appears in the list experiment Figure 8b. Bothresults together can be seen as an indicator that both inter-action techniques are rather inappropriate for complex taskslike Web browsing on the smartwatch. However, some par-ticipants remarked that they prefer the pen as they couldnot see the items in the list during the finger input condition.

Of greater relevance is the third part of the study, which isvisualized in graphs (c) and (d). The error rate visible in (c),calculated with dynamic time warping (DTW), shows thatmore precise lines can be drawn with the pen. Especiallywith increasing difficulty it is noticeable that smaller devia-tions from the templates can be achieved with the pen. Thisis also evident in the letters M and ß. A post-hoc test withBonferroni correction showed significant differences be-tween finger and stylus input on all methods except for theeasy and medium templates.

In contrast, the task completion times shown in Figure 8dindicate that the participants have drawn the letters partic-ularly quickly with the finger, whereas they spent more timewith the stylus and drew more accurately. When tracingthe rectangle templates the participants were much fasterwith the pen, even for the most difficult condition. We ex-plain this with the observation that tracing with the fingeris more difficult, as the displayed template is covered bythe finger. This results in a longer time needed for drawingand a larger error with the finger than with the stylus. Hereas well, a post-hoc test with Bonferroni correction showedsignificant differences between both input methods for alltemplates.

In general, we find that for drawing applications indirect peninput on the back of the hand is substantially faster andmore precise than direct finger input. In addition, we askedthe participants about their impressions. In general, theparticipants found the interaction on the back of their handsintuitive and pleasant. They stated that they could imagineusing such a device for drawings, web apps and games,despite the small display.

Measurement

Pointing Test: Tool vs. Time & Error

2

4

6

8

10

12

14

Err

or

[mm

]

StylusFinger

0

1

2

3

4

5

6

Tim

e [s

]

Time Error

Stylus FingerTool Type

50

100

150

200

250

300

350

DTW

Dis

tance

Painting Test: Tool vs. Error

Template

Mß

Rect Easy

Rect HardRect Medium

Stylus FingerTool Type

0

Painting Test: Tool vs. Time

Template

Mß

Rect Easy

Rect HardRect Medium

2

4

6

8

10

12

14

16

Tim

e [s

]

0

5

10

15

20

25

Tim

e [s

]

List Test: Tool vs. Time & Error

Err

or

[mm

-1]

0

2

4

6

8

10

12

14StylusFinger

MeasurementTime Error

b

a

d

c

Figure 8: Evaluation of the studyparts across all participants.

ConclusionWe proposed an interaction technique and device for draw-ing on the back of the hand for ultra small wrist-worn dis-plays. By integrating a magnet in a stylus, the position rel-ative to a magnetometer can be determined with high pre-cision. Different dimensions and magnetic field strengthswere tested with the result that a magnet with a dimensionof 7.5×20 mm and a magnetization degree of N52 achievesacceptable results on the Huawei Watch 2. Furthermore,different algorithms were evaluated regarding their accuracyin the interaction space, which showed that the 3D positiondetermination is significantly more accurate. An initial userstudy served to evaluate the field of application. It showedthat by using the stylus, more precise interactions can bemade than with the finger. Contrarily, the interaction timeincreases with the pen. However, the study shows that forapplications like drawing on a tiny screen, direct finger inputleads to higher error rates and higher task completion timesthan indirect pen input on the back of the hand.

In the future we plan to identify further interactions that areparticularly suitable for pen-based input on the back of thehand. Additional tests are necessary, e.g., to optimize theapproach for interactions like web browsing, gaming, or key-board input. We will also investigate whether this technol-ogy can be used for authentication and handwriting recogni-tion on the back of the hand.

REFERENCES[1] Tetsuya Abe, Buntarou Shizuki, and Jiro Tanaka. 2016.

Input Techniques to the Surface Around a SmartphoneUsing a Magnet Attached on a Stylus. In Proceedingsof the 2016 CHI Conference Extended Abstracts onHuman Factors in Computing Systems (CHI EA ’16).ACM, New York, NY, USA, 2395–2402. DOI:http://dx.doi.org/10.1145/2851581.2892376

[2] National Aeronautics and Space Administration. 1995.NASA-STD-3000, Man-System Integration Standards,Volume 1, Section3 : Anthropometric DimensionalData. 1 (1995). https://msis.jsc.nasa.gov

[3] Daniel Ashbrook, Patrick Baudisch, and Sean White.2011. Nenya: Subtle and eyes-free mobile input with amagnetically-tracked finger ring. Conference onHuman Factors in Computing Systems - Proceedings(05 2011), 2043–2046. DOI:http://dx.doi.org/10.1145/1978942.1979238

[4] Alex Butler, Shahram Izadi, and Steve Hodges. 2008.SideSight: Multi-"Touch" Interaction Around SmallDevices. In Proceedings of the 21st Annual ACMSymposium on User Interface Software andTechnology (UIST ’08). ACM, New York, NY, USA,201–204. DOI:http://dx.doi.org/10.1145/1449715.1449746

[5] Juan Manuel Camacho and Victor Sosa. 2013.Alternative method to calculate the magnetic field ofpermanent magnets with azimuthal symmetry. Revistamexicana de fisica E 59 (06 2013), 8–17.

[6] Ke-Yu Chen, Shwetak Patel, and Sean Keller. 2016.Finexus: Tracking Precise Motions of MultipleFingertips Using Magnetic Sensing. 1504–1514. DOI:http://dx.doi.org/10.1145/2858036.2858125

[7] Xiang ’Anthony’ Chen, Tovi Grossman, and GeorgeFitzmaurice. 2014. Swipeboard: A Text EntryTechnique for Ultra-small Interfaces That SupportsNovice to Expert Transitions. In Proceedings of the27th Annual ACM Symposium on User InterfaceSoftware and Technology (UIST ’14). ACM, New York,NY, USA, 615–620. DOI:http://dx.doi.org/10.1145/2642918.2647354

[8] Mark D. Dunlop, Andreas Komninos, and NaveenDurga. 2014. Towards High Quality Text Entry onSmartwatches. In CHI ’14 Extended Abstracts onHuman Factors in Computing Systems (CHI EA ’14).ACM, New York, NY, USA, 2365–2370. DOI:http://dx.doi.org/10.1145/2559206.2581319

[9] Liming Fan, Chong Kang, Xiaojun Zhang, QuanZheng, and Ming Wang. 2016. An efficient method fortracking a magnetic target using scalar magnetometerarray. SpringerPlus 5 (12 2016). DOI:http://dx.doi.org/10.1186/s40064-016-2170-0

[10] Jun Gong, Zheer Xu, Qifan Guo, Teddy Seyed,Xiang ’Anthony’ Chen, Xiaojun Bi, and Xing-DongYang. 2018. WrisText: One-handed Text Entry onSmartwatch Using Wrist Gestures. In Proceedings ofthe 2018 CHI Conference on Human Factors inComputing Systems (CHI ’18). ACM, New York, NY,USA, Article 181, 14 pages. DOI:http://dx.doi.org/10.1145/3173574.3173755

[11] Jun Gong, Xing-Dong Yang, and Pourang Irani. 2016.WristWhirl: One-handed Continuous Smartwatch Inputusing Wrist Gestures. 861–872. DOI:http://dx.doi.org/10.1145/2984511.2984563

[12] Mitchell Gordon, Tom Ouyang, and Shumin Zhai.2016. WatchWriter: Tap and Gesture Typing on aSmartwatch Miniature Keyboard with StatisticalDecoding. In Proceedings of the 2016 CHI Conferenceon Human Factors in Computing Systems (CHI ’16).ACM, New York, NY, USA, 3817–3821. DOI:http://dx.doi.org/10.1145/2858036.2858242

[13] Aakar Gupta and Ravin Balakrishnan. 2016. DualKey:Miniature Screen Text Entry via Finger Identification. InProceedings of the 2016 CHI Conference on HumanFactors in Computing Systems (CHI ’16). ACM, NewYork, NY, USA, 59–70. DOI:http://dx.doi.org/10.1145/2858036.2858052

[14] Jonggi Hong, Seongkook Heo, Poika Isokoski, andGeehyuk Lee. 2015. SplitBoard: A Simple Split SoftKeyboard for Wristwatch-sized Touch Screens. InProceedings of the 33rd Annual ACM Conference onHuman Factors in Computing Systems (CHI ’15).ACM, New York, NY, USA, 1233–1236. DOI:http://dx.doi.org/10.1145/2702123.2702273

[15] Hamed Ketabdar, Mehran Roshandel, and Kamer AliYüksel. 2010. Towards Using Embedded MagneticField Sensor for Around Mobile Device 3D Interaction.In Proceedings of the 12th International Conference onHuman Computer Interaction with Mobile Devices andServices (MobileHCI ’10). ACM, New York, NY, USA,153–156. DOI:http://dx.doi.org/10.1145/1851600.1851626

[16] Wolf Kienzle and Ken Hinckley. 2013. WritingHandwritten Messages on a Small Touchscreen. InProceedings of the 15th International Conference onHuman-computer Interaction with Mobile Devices and

Services (MobileHCI ’13). ACM, New York, NY, USA,179–182. DOI:http://dx.doi.org/10.1145/2493190.2493200

[17] Jarrod Knibbe, Diego Martinez Plasencia, ChristopherBainbridge, Chee-Kin Chan, Jiawei Wu, ThomasCable, Hassan Munir, and David Coyle. 2014.Extending Interaction for Smart Watches: EnablingBimanual Around Device Control. In CHI ’14 ExtendedAbstracts on Human Factors in Computing Systems(CHI EA ’14). ACM, New York, NY, USA, 1891–1896.DOI:http://dx.doi.org/10.1145/2559206.2581315

[18] Luis A. Leiva, Alireza Sahami, Alejandro Catala, NielsHenze, and Albrecht Schmidt. 2015. Text Entry on TinyQWERTY Soft Keyboards. In Proceedings of the 33rdAnnual ACM Conference on Human Factors inComputing Systems (CHI ’15). ACM, New York, NY,USA, 669–678. DOI:http://dx.doi.org/10.1145/2702123.2702388

[19] S. O. H. Madgwick, A. J. L. Harrison, and R.Vaidyanathan. 2011. Estimation of IMU and MARGorientation using a gradient descent algorithm. In 2011IEEE International Conference on RehabilitationRobotics. 1–7. DOI:http://dx.doi.org/10.1109/ICORR.2011.5975346

[20] Jess McIntosh, Paul Strohmeier, Jarrod Knibbe,Sebastian Boring, and Kasper Hornbæk. 2019.Magnetips: Combining Fingertip Tracking and HapticFeedback for Around-Device Interaction. InProceedings of the 2019 CHI Conference on HumanFactors in Computing Systems (CHI ’19). ACM, NewYork, NY, USA, Article 408, 12 pages. DOI:http://dx.doi.org/10.1145/3290605.3300638

[21] Ian Oakley and Doyoung Lee. 2014. Interaction on theEdge: Offset Sensing for Small Devices. InProceedings of the SIGCHI Conference on HumanFactors in Computing Systems (CHI ’14). ACM, NewYork, NY, USA, 169–178. DOI:http://dx.doi.org/10.1145/2556288.2557138

[22] Masa Ogata and Michita Imai. 2015. SkinWatch: skingesture interaction for smart watch. In AH ’15.

[23] Stephen Oney, Chris Harrison, Amy Ogan, and JasonWiese. 2013. ZoomBoard: A Diminutive Qwerty SoftKeyboard Using Iterative Zooming for Ultra-smallDevices. In Proceedings of the SIGCHI Conference onHuman Factors in Computing Systems (CHI ’13).ACM, New York, NY, USA, 2799–2802. DOI:http://dx.doi.org/10.1145/2470654.2481387

[24] Farshid Salemi Parizi, Eric Whitmire, and Shwetak N.Patel. 2019. AuraRing: Precise Electromagnetic FingerTracking. IMWUT 3 (2019), 150:1–150:28.

[25] Simon Perrault, Eric Lecolinet, James Eagan, andYves Guiard. 2013. WatchIt: Simple gestures andeyes-free interaction for wristwatches and bracelets.1451–1460. DOI:http://dx.doi.org/10.1145/2470654.2466192

[26] Henning Pohl, Dennis Stanke, and Michael Rohs.2016. EmojiZoom: Emoji Entry via Large OverviewMaps. In Proceedings of the 18th InternationalConference on Human-Computer Interaction withMobile Devices and Services (MobileHCI ’16). ACM,New York, NY, USA, 510–517. DOI:http://dx.doi.org/10.1145/2935334.2935382

[27] Ryan Qin, Suwen Zhu, Yu-Hao Lin, Yu-Jung Ko, andXiaojun Bi. 2018. Optimal-T9: An Optimized T9-likeKeyboard for Small Touchscreen Devices. In

Proceedings of the 2018 ACM InternationalConference on Interactive Surfaces and Spaces (ISS’18). ACM, New York, NY, USA, 137–146. DOI:http://dx.doi.org/10.1145/3279778.3279786

[28] Gabriel Reyes, Jason Wu, Nikita Juneja, MaximGoldshtein, W. Keith Edwards, Gregory D. Abowd, andThad Starner. 2018. SynchroWatch: One-HandedSynchronous Smartwatch Gestures Using Correlationand Magnetic Sensing. Proc. ACM Interact. Mob.Wearable Ubiquitous Technol. 1, 4, Article 158 (Jan.2018), 26 pages. DOI:http://dx.doi.org/10.1145/3161162

[29] Stefan Schneegaß and Alexandra Voit. 2016.GestureSleeve: using touch sensitive fabrics forgestural input on the forearm for controllingsmartwatches. 108–115. DOI:http://dx.doi.org/10.1145/2971763.2971797

[30] Maximilian Schrapel, Max-Ludwig Stadler, and MichaelRohs. 2018. Pentelligence: Combining Pen Tip Motionand Writing Sounds for Handwritten Digit Recognition.In Proceedings of the 2018 CHI Conference on HumanFactors in Computing Systems (CHI ’18). Associationfor Computing Machinery, New York, NY, USA, ArticlePaper 131, 11 pages. DOI:http://dx.doi.org/10.1145/3173574.3173705

[31] Tomoki Shibata, Daniel Afergan, Danielle Kong,Beste F. Yuksel, Scott MacKenzie, and Robert J.K.Jacob. 2016. Text Entry for Ultra-Small TouchscreensUsing a Fixed Cursor and Movable Keyboard. InProceedings of the 2016 CHI Conference ExtendedAbstracts on Human Factors in Computing Systems(CHI EA ’16). ACM, New York, NY, USA, 3770–3773.DOI:http://dx.doi.org/10.1145/2851581.2890230

[32] Srinath Sridhar, Anders Markussen, Antti Oulasvirta,Christian Theobalt, and Sebastian Boring. 2017.WatchSense: On- and Above-Skin Input Sensingthrough a Wearable Depth Sensor. 3891–3902. DOI:http://dx.doi.org/10.1145/3025453.3026005

[33] Paul Strohmeier, Jesse Burstyn, and Roel Vertegaal.2015. Effects of Display Sizes on a Scrolling TaskUsing a Cylindrical Smartwatch. In Proceedings of the17th International Conference on Human-ComputerInteraction with Mobile Devices and Services Adjunct(MobileHCI ’15). ACM, New York, NY, USA, 846–853.DOI:http://dx.doi.org/10.1145/2786567.2793710

[34] Malte Weiss, Chat Wacharamanotham, Simon Voelker,and Jan Borchers. 2011. FingerFlux: Near-surfaceHaptic Feedback on Tabletops. In Proceedings of the24th Annual ACM Symposium on User InterfaceSoftware and Technology (UIST ’11). ACM, New York,NY, USA, 615–620. DOI:http://dx.doi.org/10.1145/2047196.2047277

[35] Daniel Wigdor, Clifton Forlines, Patrick Baudisch, JohnBarnwell, and Chia Shen. 2007. Lucid Touch: ASee-through Mobile Device. In Proceedings of the 20thAnnual ACM Symposium on User Interface Softwareand Technology (UIST ’07). Association for ComputingMachinery, New York, NY, USA, 269âAS278. DOI:http://dx.doi.org/10.1145/1294211.1294259

[36] Pui Wong, Zhu Kening, and Hongbo Fu. 2017.fingerT9 : leveraging thumb-to-finger interaction forone-handed text entry on smartwatches. 1–3. DOI:http://dx.doi.org/10.1145/3132787.3132806

[37] Po-Chen Wu, Robert Wang, Kenrick Kin, ChristopherTwigg, Shangchen Han, Ming-Hsuan Yang, and

Shao-Yi Chien. 2017. DodecaPen: Accurate 6DoFTracking of a Passive Stylus. Association forComputing Machinery, New York, NY, USA.

[38] Haijun Xia, Tovi Grossman, and George Fitzmaurice.2015. NanoStylus: Enhancing Input on Ultra-SmallDisplays with a Finger-Mounted Stylus. In Proceedingsof the 28th Annual ACM Symposium on User InterfaceSoftware & Technology (UIST ’15). ACM, NewYork, NY, USA, 447–456. DOI:http://dx.doi.org/10.1145/2807442.2807500

[39] Robert Xiao, Gierad Laput, and Chris Harrison. 2014.Expanding the Input Expressivity of Smartwatcheswith Mechanical Pan, Twist, Tilt and Click. InProceedings of the SIGCHI Conference on HumanFactors in Computing Systems (CHI ’14). ACM, NewYork, NY, USA, 193–196. DOI:http://dx.doi.org/10.1145/2556288.2557017

[40] Xin Yi, Chun Yu, Weijie Xu, Xiaojun Bi, and YuanchunShi. 2017. COMPASS: Rotational Keyboard onNon-Touch Smartwatches. In Proceedings of the 2017CHI Conference on Human Factors in ComputingSystems (CHI ’17). ACM, New York, NY, USA,705–715. DOI:http://dx.doi.org/10.1145/3025453.3025454

[41] Sang Ho Yoon, Ke Huo, and Karthik Ramani. 2016.TMotion: Embedded 3D Mobile Input Using MagneticSensing Technique. In Proceedings of the TEI ’16:Tenth International Conference on Tangible,Embedded, and Embodied Interaction (TEI ’16). ACM,New York, NY, USA, 21–29. DOI:http://dx.doi.org/10.1145/2839462.2839463