（School of Instrument Science and Engineering，Southeast University，Nanjing 210096，China）

Evaluating effectiveness of four-parameter compound tactons for conveying urban navigation information

Jia Zhen Li Jianqing Xiao Yong

（School of Instrument Science and Engineering，Southeast University，Nanjing 210096，China）

To guide pedestrians to navigate in a strange city and reduce the cognitive overload suffered when walking，thermal change direction，vibration stimulation direction，intensity variation and body location are employed to construct four-parameter compound tactons．They are mapped to four different types of navigation message：route attribute，intersection type，distance and heading direction．One psychological experiment was conducted．The derived confusion matrices were used to investigate recognition rates and information transfer for compound tactons，and nonparameter tests were employed to analyze the effect of each parameter on the number of correct responses．Experimental results show that the overall identification rate for fourparameter tactons is 88．72%by using different tactile parameters，and 19．64 icons can be identified reliably in all32 tactile icons according to the information transfer value．Thermal changes can be an effective supplement to vibrotactile icons．This suggests that compound tactons will be a promising method of conveying complex information when navigating in a virtual or real urban environment．

tactile icons；navigation information；thermal display；vibrotactile feedback

Nowadays，more frequentmobility activitiesare taking place in the modern and fast-paced society．One challenge for pedestrians is orientation in unknown spaces．However，it is difficult to navigate in a strange place．Even if there are fixed traffic signs，they cannot be seen when this place is very crowded，and are unavailable to blind people．Most navigation systems convey spatial information through the visual or audio channel，by which the information can nearly be accessed everywhere．However for the blind，they are neither hand-free nor an effective solution．Using visual and auditory channels can always distract the users’attention，and the information may be overwhelm ing due to sensory overload．According to W ickens’multiple resource theory（MRT）［1］，allocating information from one overloaded sensemodality to anothermodality can reduce workload．Therefore，the alternative tactile channel can be employed to convey navigation information．

There have been a number of studies on using vibrotactile feedback for navigation．Such tactile displays often took the form of vests or belts．The ActiveBelt invented by Tsukada et al．［2］was a waist belt equipped w ith eight vibrators，producing vibrotactile stimuli around the waist to“point”into a horizontal direction．W ith a sim ilar waist belt，Pielot et al．［3］performed several studies on how pedestrians received navigation information in the form of vibrotactile feedback．Distance presentation can bemore beneficial for spatial navigation，although direction information alone is sufficient to reach a turn．M c-Daniel et al．［4］presented intimate，personal，and social interpersonal distances to the blind by using tactile rhythms．Straub et al．［5］designed a vibrotactile waist belt to provide distance information for the pedestrians．By using a tactile torso display，Pielot et al．［6］presented the location and spatial distance of several people in a fastpaced 3D multiplayer game．Tactile icons（“tactons”），first proposed by Brewster et al．［7］，are classified as oneelement，compound，inherited，and transformational tactons．As structured vibrotactilemessages，transformational tactonswere later employed to provide navigation feedback．Lin et al．［8］proposed a pedestrian navigation system of conveying navigation information to the sighted user by using tactons．Different from transformational tactons，compound tactons are a sequence of two ormore one-element tactile icons．However，few studies have used compound tactons to present non-visual information．In Ref．［9］，we designed three-parameter compound tactons to encode urban navigation information，and found that the overall identification rate was rather high at 92．65%by using three vibrotactile parameters．This indicates that compound tactons are feasible to convey complex information．

In this paper，four-parameter tactons are employed to convey multi-dimensional navigation information in the form of thermal and vibrotactile stimuli．Based on the sem i-infinite bodymodel，the thermal responsewithin the skin is investigated theoretically．One psychophysical experiment is performed to evaluate the effectiveness ofcompound tactons for communicating navigation messages．Experimental results indicate that compound tactons have good performance in conveying non-visual information，and thermal changes can be an effective supplement to vibrotactile feedback．

1 Thermal Response w ithin the Skin

The semi-infinite body model is a reasonable initial choice tomodel the contact between the finger and an object，and the model-based thermal display was proved to be effective in simulating differentmaterials［10］．In previous research，thermal displays always served as a constant temperature source and maintained the interface temperature．However，the detection of thermal changes is dependentmore on the stimulus change rate than the actual extent of the change itself（absolute temperature）［11］．Therefore，manipulating both thermal change rate and thermal intensity can perceptually create distinct stimuli．We propose that the thermal display can vary linearly from the initial skin temperature Ts，0，and then the interface temperature Tcis

where k is a constant rate of thermal change，and t is the time．Based on the semi-infinite bodymodel，the temperature profileswithin the skin（for x＞0）can be calculated by

where Ts，0is the initial skin temperature；αsis the thermal diffusivity of the skin．The heat flux conducted out of the skin during contact can be solved by

where Qsis the heat flux conducted out of the skin during contact；λsis the thermal conductivity of the skin．The heat flux changes w ithin the skin will result in different thermal sensations in the human central nervous system．This significant difference may be in favor of identifying and discrim inating variousmessages．

2 Experimental Setup and Icon Design

2.1 Tactile system construction

The experimental apparatus mainly consists of navigation information acquiring and display modules，a kernel control module，vibrotactile feedback module，and thermal display module，aswell as the power supply．As shown in Fig．1，the information acquiring unit is a NRF24L01 w irelessmodule，which can receive the transm itted navigation information w irelessly，and send it to the kernel control module．An electronic compass is in charge of collecting the orientation information from the user himself．The kernel control module is a STM 32VET6 controller，whose functions are decoding navigation information and encoding tactile language in the form of vibrotactile and thermal stimuli．The information display module is a TFT-LCD screen，displaying the received navigation information in real time．The vibrotactile feedback module includes the vibrotactile driving circuit and a waist belt，of which four vibrating motors are attached to the inside of the belt，and the driving circuit enables vibrators to produce vibrotactile stimulus．The thermal display module includes the thermal tactile driving circuit，the temperature sensor circuit and an armband．A Peltier w ith a heat sink is embedded into the armband．The sensor circuitmeasures the temperature of the Peltier surface and sends its values to the kernel controlmodule in real time，and the driving circuit enables the Peltier to generate thermal stimulus．Besides，a backpack is chosen to carry the experimental apparatus．Fig．2 presents the prototype of the wearable tactile navigation system．

Fig.1 Assembly photograph of the controller board

Fig.2 Prototype of the wearable tactile navigation system．（a）Front view；（b）Lateral view

2.2 Design of four-param eter tactons

Follow ing the design principle of three-parameter tactons in Ref．［9］，the four-parameter tactons were structured by adding another piece of information：the“Route Attribute”of a hypothetical navigation message，w ith the purpose to reduce navigation errors of walking in the urban environment．The route attribute can be either“deviation”or“on route”．However，“destination”of the intersection type is removed in this case，as it appears only when reaching the destination in the real scenario．Therefore，the intersection type will be“crossing”or“l(fā)andmark”．As a result，this w ill generate 32 different message types．The extra thermal parameter chosen is the thermal change direction，as it is proved to be particularly salient［13］．This direction has two levels：warm ing and cooling．The chosen initial skin temperature is 30℃，a neutral skin temperature also used in Ref．［12］，w ith stimuliwarm_ed and cooled from this temperature．In previous works［1314］，1℃／s was reported to be sufficient to produce detectable sensations in ideal conditions，and faster 3℃／s could create perceptually more distinct stimuli．The thermal change rate adopted in this paper for warm ing or cooling was 3℃／s，w ith the action time lasting 3 s．Warm ing is used to represent on-routemessages，because warm stimuliare generally more preferred than cold one［15］．Cooling is then mapped to deviation messages．Themapping relationships between icon and message type are illustrated in Fig．3．For the“distance”and“heading direction”，their mappings between icon and message type keep unchanged［9］．

Fig.3 Encoding categories of route attributes，intersection types，distances and heading directions

3 Evaluation of Four-Parameter Tactons

The experimental results in Ref．［9］indicated that the mean identification rate for three-parameter compound tactons is almost10%higher than that for three-parameter transformational tactons［16］，and subjects can identify almost 18 icons correctly of all 24 tactile icons．Therefore，we carried out an experiment to investigate subjects’ability to discrim inate four-parameter compound thermal＋vibrotactile icons．

3.1 Design and procedure

Sixteen normal healthy adults participated in this experiment（13 males and 3 females），aged between 23 and 29 yearsw ith an average age of25．4．They were all students from Southeast University，and none of them had participated in the previous experiment in Ref．［9］．The subjects were first trained to recognize the four-parameter tactons prior to the experiment．The concept of iconswas introduced to them，and the mapping relationships between vibrotactile parameters and attributes of the navigationmessageswere explained．Then they were allowed to sense each icon several times in 15 m in，by clicking the buttons on the training interface．During this time，they had to associate the icons w ith theirmeanings．After becom ing fam iliarwith the icons，they took part in 32 tasks like those from the experiment itself as training in how to use the testing interface．Headphoneswere worn to block any noise from the vibrators．

All 32 icons were presented four times random ly，resulting in 128 trials for each subject in the experiment．Therefore，there were a total of 2 048 trials（i．e．，16 subjects×2 thermal change directions×2 stimulation directions×2 intensity variations×4 body locations×4 presentations）．In each trial，one icon was repeated two times with a one-second pause between repetitions．While the icon was being presented，the subjects had to identify all four attributes encoded in the icon，by selecting the corresponding radio buttons on the testing interface．They did not have to wait until the icon had been presented two times before responding if they had already decided．Then，they could subm it their responses to carry out the next trial．All the experiment interfaces were built in a Visual C＋＋environment．

3.2 Results and discussion

Fig．4 presents the confusionmatrix of subjects’responses for the main session．The overallmean accuracy for four-parameter tactons is 88．72%．From the confusion matrix，the information transfer Itfrom the stimulus set to the subject can be calculated by［17］

where Pmis the probability of stimulus m；Pnis the probability of response n；Pmnis the joint probability of stimulus m and response n．According to the relation N＝2It，The Itcan be considered to indicate the number N of categories perceived by the subject．After calculation，the Itvalue is 4．30 bits，and 19．64 icons can，therefore，be distinguished reliably．Furthermore，the mean accuracy for the four thermal＋vibrotactile parameters is 95．85% for thermal change direction，97．61%for stimulation di-rection，95．8%for intensity variation，and 98．63%for body location．The confusionmatrix for each parameter is shown in Fig．5．

Fig.4 Confusion matrix in themain session

Shapiro-W ilk tests showed that the data for four-parameter tactons violated the assumption of a normal distribution，so non-parametric testswere used，specifically W ilcoxon signed-rank tests for the effects of thermal change direction，stimulation direction，and intensity variation，and Friedman’s test for the effect of body location．Post hoc pairw ise comparisons follow ing a significant Friedman’s testwere conducted using theW ilcoxon tests．Friedman’s analysis of variance by ranks showed no effect of body location on the number of correct responses（χ2（3）＝4．178，p＝0．194），w ith means of 98．44%，98．83%，99．61%and 97．66%for the right，back，left and front，respectively．A W ilcoxon test showed that，the thermal change direction had a significant effect on the correct response counts，as subjects identified significantly more warm icons than cold icons（Z＝－3．189，p＝0．001）．Themean recognition rates were 92．97%for cooling and 98．73%for warm ing．However，there were no significantmain effects of either stimulation direction（Z＝－0．159，p＝0．874）or intensity variation（Z＝－0．361，p＝0．718）on the number of correct responses．Mean recognition rates for stimulation direction were 97．75%for clockw ise（CW）and 97．46%for counterclockw ise（CCW），while mean recognition rates for intensity variation were 94．92%for decreasing and 96．68%for increasing．Fig．6 shows themean identification rates of all four-parameter tactile icon types．

Compared w ith three-parameter compound tactons in Ref．［9］，four-parameter tactons have a lowermean identification rate at 88．72%，w ith 95．85%accuracy for thermal change direction，97．61%accuracy for stimulation direction，95．8%accuracy for intensity variation，and 98．63%accuracy for body location．Therefore，it appears that presenting thermal and vibrotactile stimuli sequentially does not significantly hinder interpretation of each other，and thermal changesmay be an effective supplement to vibrotactile icons．When increasing the amount of information input，the amount of informationtransm itted for four-parameter tactons increases slightly，w ith 4．30 bits corresponding to 19．64 icons．In other words，of all 32 tactile icons，around 19 to 20 icons could be identified reliably．A non-parametric test indicated that，cold icons were significantly more difficult to identify than the warm icons．The reason is that the heat sink cannot efficiently dissipate the accumulated heat，which w ill distort subjects’judgmentwhen the Peltier actually presents a cold stimulus．

Fig.5 Confusionmatrix for each parameter．（a）Thermal change direction；（b）Stimulation direction；（c）Intensity variation；（d）Body location

Fig.6 Mean identification rates for all four icon types

4 Conclusion

This paper has conducted the evaluation of four-parameter thermal and vibrotactile icons，and the thermal response w ithin the skin is studied on the basis of the typical sem i-infinite body model．To evaluate the effectiveness of multi-dimensional compound tactons for navigation information displays，one psychophysical experiment is carried out，reporting that four-parameter tactons have amean identification rate at 88．72%from themain confusion matrix，and about 19 to 20 icons can be identified reliably from all 32 tactile icons based on the information transfer value．It can be concluded that identifying four types of navigation information from the tactons is easy，and does not seem to produce confusion．A necessary next step is testing identification while the user iswalking in a virtual or real urban environment．These initial results show compound tactons to be sim ilarly effective in conveying multidimensional information as earcons and transformational tactons．

However，a pedestrian hasmore choices of orientation in the urban environment that need higher accuracy．Therefore，adding another four ordinal directions around the waistw ill be considered．To elim inate the significant differences between warm and cold stimuli，two Peltiers instead of one w ill be employed to cover a larger stimulusarea［13］．A lthough the identification accuracy for compound tactons is distinctly high，the total duration required for perceiving sequential messages may be too long，making the information transfer rate very slow．Therefore，compound tactons can be combined w ith transformational tactons to convey navigation information．

［1］Wickens C D．Multiple resources and performance prediction［J］．Theoretical Issues in Ergonomics Science，2002，3（2）：159- 177．

［2］Tsukada K，Yasumura M．Activebelt：belt-type wearable tactile display for directional navigation［C］／／ACM International Joint Conference on Pervasive and Ubiquitous Computing．Nottingham，UK，2004：384- 399．

［3］Pielot M，Henze N，Heuten W，et al．Evaluation of continuous direction encoding w ith tactile belts［C］／／3rd InternationalWorkshop on Haptic and Audio Interaction Design．Jyvaskyla，F(xiàn)inland，2008：1- 10．

［4］McDaniel T，Kirshna S，Colbry D．Using tactile rhythm to convey interpersonal distances to individuals who are blind［C］／／Proceedings of the Conference on Human Factors in Computing Systems．Boston，MA，USA，2009：4669- 4674．

［5］Straub M，Riener A，F(xiàn)erscha A．Distance encoding in vibro-tactile guidance cues［C］／／6th Annual International Conference on Mobile and Ubiquitous Systems：Computing，Networking and Services．Toronto，Canada，2009：1- 2．

［6］Pielot M，Krull O，Boll S．Where ismy team？Supporting collaboration and situation awareness w ith tactile displays［C］／／Proceedings of the Conference on Human Factors in Computing Systems．Atlanta，GA，USA，2010：1705- 1714．

［7］Brewster S A，Brown L M．Tactons：structured tactile messages for non-visual information display［C］／／Proceedings of the 5th Conference on Australasian User Interface．Darlinghurst，Australia，2004：15- 23．

［8］Lin M，Cheng Y，Yu W．Using tactons to provide navigation cues in pedestrian situations［C］／／Proceedings of the 5th Nordic Conference on Human-Computer Interaction：Building Bridges．Lund，Sweden，2008：507- 510．

［9］Jia Z，Li JQ，Geng W P．Design of 3-D compound tactons for navigation information display［C］／／5th International Workshop on Computer Science and Engineering．Moscow，Russia，2015：776- 780．

［10］Ho H N，Jones L A．Development and evaluation of a thermal display for material identification and discrim ination［J］．ACM Transactions on Applied Perception，2007，4（2）：1- 24．

［11］Stevens JC．The psychology of touch［M］．Hillsdale，NJ，USA：Law rence Erlbaum，1991：61- 90．

［12］Richter H，Hausen D，Osterwald S，et al．Reproducing materials of virtual elements on touchscreens using supplemental thermal feedback［C］／／14th ACM International Conference on Multimodal Interaction．Santa Monica，CA，USA，2012：385- 392．

［13］W ilson G，Brewster S，Halvey M，et al．Thermal icons：evaluating structured thermal feedback formobile interaction［C］／／14th International Conference on Human-Computer Interaction with Mobile Devices and Services．San Francisco，CA，USA，2012：309- 312．

［14］Nakashige M，Kobayashi M，Suzuki Y．“HiyaAtsu”media：augmenting digitalmedia w ith temperature［C］／／Proceedings of the Conference on Human Factors in Computing Systems．Boston，MA，USA，2009：3181- 3186．

［15］Lee W，Lim Y K．Thermo-message：exploring the potential of heat as a modality of peripheral expression［C］／／Proceedings of the Conference on Human Factors in Computing Systems．Atlanta，GA，USA，2010：4231- 4236．

［16］AzadiM，Jones L A．Evaluating vibrotactile dimensions for the design of tactons［J］．IEEE Transactions on Haptics，2014，7（1）：14- 23．

［17］AzadiM，Jones L．Identification of vibrotactile patterns：building blocks for tactons［C］／／IEEE World Haptics Conference．Daejeon，Korea，2013：347- 352．

表達(dá)城市導(dǎo)航信息的四參數(shù)復(fù)合式觸覺圖標(biāo)效能評估

賈 貞 李建清 肖 勇

（東南大學(xué)儀器科學(xué)與工程學(xué)院，南京210096）

為了幫助行人在陌生的城市中尋路和降低行走時的認(rèn)知負(fù)擔(dān)，利用溫度的變化方向、振動的刺激方向、強(qiáng)度變化以及身體位置4個觸覺刺激參數(shù)構(gòu)造了復(fù)合式的觸覺圖標(biāo)，分別映射為4種不同的導(dǎo)航信息：路線屬性、路口類型、距離和行進(jìn)方向．開展了一組心理物理學(xué)實驗，利用模糊矩陣研究參與者對觸覺圖標(biāo)的識別正確率和信息傳輸能力，并使用非參數(shù)檢驗分析了4個觸覺參數(shù)對正確響應(yīng)數(shù)量的影響．實驗結(jié)果表明：使用不同觸覺參數(shù)構(gòu)造出的復(fù)合式觸覺圖標(biāo)，平均識別正確率可達(dá)到88．72%；根據(jù)計算出的信息傳輸值，在所有的32個觸覺圖標(biāo)中共有19．64個圖標(biāo)能夠被可靠地辨識；溫度變化可作為振動觸覺圖標(biāo)的一種有效補(bǔ)充．在以后的虛擬或真實城市環(huán)境中導(dǎo)航中，復(fù)合式觸覺圖標(biāo)將是一種有潛力的傳遞這類復(fù)雜信息的方法．

觸覺圖標(biāo)；導(dǎo)航信息；熱顯示；振動觸覺反饋

TP391

10．3969／j．issn．1003－7985．2015．03．011

2014-12-31．

Biographies：Jia Zhen（1987—），male，graduate；Li Jianqing（corresponding author），male，doctor，professor，ljq＠seu．edu．cn．

s：The Natural Science Foundation of Jiangsu Province（No．BK2012560），the Research Fund for the Doctoral Program of Higher Education of China（No．20130092110060）．

：Jia Zhen，Li Jianqing，Xiao Yong．Evaluating effectiveness of four-parameter compound tactons for conveying urban navigation information［J］．Journal of Southeast University（English Edition），2015，31（3）：363- 368．

10．3969／j．issn．1003－7985．2015．03．011