key: cord-1046718-ncjolikm
authors: Guarese, Renan; Andreasson, Pererik; Nilsson, Emil; Maciel, Anderson
title: Augmented Situated Visualization Methods towards Electromagnetic Compatibility Testing
date: 2020-10-15
journal: Comput Graph
DOI: 10.1016/j.cag.2020.10.001
sha: 8b9064423dc5f4357ef5dd7863cb02595e6016cc
doc_id: 1046718
cord_uid: ncjolikm

In electrical engineering, hardware experts often need to analyze electromagnetic radiation data to detect any external interference or anomaly. The field that studies this sort of assessment is called electromagnetic compatibility (EMC). As a way to support EMC analysis, we propose the use of Augmented Situated Visualization (ASV) to supply professionals with visual and interactive information that helps them to comprehend that data, however situating it where it is most relevant in its spatial context. Users are able to interact with the visualization by changing the attributes being displayed, comparing the overlaps of multiple fields, and extracting data, as a way to refine their search. The solutions being proposed in this work were tested against each other in comparable 2D and 3D interactive visualizations of the same data in a series of data-extraction assessments with users, as a means to validate the approaches. Results exposed a correctness-time trade-off between the interaction methods. The hand-based techniques (Hand Slider and Touch Lens) were the least error-prone, being near to half as error-inducing as the gaze-based method. Touch Lens also performed as the least time-consuming method, taking in average less than half of the average time required by the others. For the visualization methods tested, the 2D ray casts presented a higher usability score and lesser workload index than the 3D topology view, however exposing over two times the error ratio. Ultimately, this work exposes how AR can help users to have better performances in a decision-making context, particularly in EMC related tasks, while also furthering the research in the ASV field.

In electrical engineering, hardware experts often need to analyze electromagnetic radiation data to detect any external interference or anomaly. The field that studies this sort of assessment is called electromagnetic compatibility (EMC). As a way to support EMC analysis, we propose the use of Augmented Situated Visualization (ASV) to supply professionals with visual and interactive information that helps them to comprehend that data, however situating it where it is most relevant in its spatial context. Users are able to interact with the visualization by changing the attributes being displayed, comparing the overlaps of multiple fields, and extracting data, as a way to refine their search. The solutions being proposed in this work were tested against each other in comparable 2D and 3D interactive visualizations of the same data in a series of data-extraction assessments with users, as a means to validate the approaches. Results exposed a correctness-time trade-off between the interaction methods. The hand-based techniques (Hand Slider and Touch Lens) were the least error-prone, being near to half as error-inducing as the gaze-based method. Touch Lens also performed as the least time-consuming method, taking in average less than half of the average time required by the others. For the visualization methods tested, the 2D ray casts presented a higher usability score and lesser workload index than the 3D topology view, however exposing over two times the error ratio. Ultimately, this work exposes how AR can help users to have better performances in a decision-making context, particularly in EMC related tasks, while also furthering the research in the ASV field. c 2020 Elsevier B.V. All rights reserved. 1 Concerning hardware testing, the activity of electronic and 2 electrical devices can be highly affected by external sources, 3 such as the frequency components of electromagnetic waves 4 emitted in natural lightning, fluorescent lights, computers, and 5 Fig. 1 . Electromagnetic Compatibility (EMC) is the study con-10 cerned with the design of electronic systems such that interfer-11 ence from or to that system will be minimized, in order not to 12 affect any of its surroundings. A system can be considered elec- 13 tromagnetically compatible with its environment if it satisfies 14 three criteria [1] :

1. It does not cause interference with other systems.

tion [4] . 23 Hereupon, this paper contributes with a review and a de- 24 sign that suggests the use of Augmented Situated Visualiza-25 tion (ASV), as well as novel interaction methods for extraction, 26 while analyzing data in the EMC field. It is particularly aimed at 27 easing decision-making according to Balleine's definition [5] . 28 Alongside an AR Optical See-Through Head-Mounted Display

(HMD) to provide in situ information regarding the EM fields 30 in the user vicinity, ASV can aid users to perceive the data with-31 out exhaustively exploring it or making a mental translation of 32 it from a 2D map perspective. In such a way, this paper also 33 contributes with a task-based study to assess user performance 34 in augmented situated visualization of electromagnetic fields.

In scientific visualization, it is customary to address novel 37 data visualization and rendering methods in a generic manner, 38 in order not to focus on any particular platforms or display 39 paradigms. Several works dealing with 3D visualization of data 40 in different domains are presented in this fashion [6, 7, 8] . trary to these, the current work focuses on the Augmented Sit-42 uated Visualization paradigm, with the aim to expose optimal 43 ways to display and interact with EMC information located in 44 situ. 45 2. 1. EMC visualization systems 46 In 2018, Sato et al. [9] developed a method to measure and 47 display the intensity of 3D EMFs in a tablet device. It was 48 possible to visualize a 3D distribution of the field measured in 49 real-time by a dosimeter using markers to position the data ac-50 cording to the actual device being measured. Their displayed 51 data, however, is quite discrete, failing to expose the continuity 52 of a 3D field, likely due to the imprecision of their measurement 53 method. In the same year, Isrie et al. [10] demonstrated a data 54 acquisition system that displays readings from a situated power 55 sensor using the current GPS location in a heads up display. 56 Their approach allows users to move along great distances and 57 still have access to the data. The data displayed in their work 58 is entirely made of 2D graphs, not properly situated in the sur-59 rounding area, which might cause users to misinterpret the pre-60 cise location of the readings. Neither of these works presented 61 any user studies. In their 2019 study, Rioult et al. [11] demonstrated an EMC 63 scanning and visualization system aimed at providing fast read-64 ings in confined and even remote environments, using a com-65 pact portable device, depicted in Fig. 2 . The device is capable 66 of measuring electromagnetic radiations as well as presenting 67 them in AR, being situated in loco as 2D grids. Their work fo-68 cuses solely on relatively small scale situations, requiring the 69 environment to be manually scanned. Again, no user studies 70 were presented to evaluate system usability. In order to ad-71 dress the aforementioned limitations, the present paper focuses 72 on exposing EMC data in a more continuous and precise man-73 ner, maintaining the 3D topology of the fields radiated from the 74 tested devices. Every piece of collected data will also be spa-75 tially situated around the hardware being analyzed, preserving 76 the ASV paradigm. 

In 2016, a study by Willet et al. [12] brought to light the ben-79 efits, trade-offs, and some linguistic definitions regarding Infor-80 mation Visualization in AR. Willet revises different application 81 works from the literature and semantically defines and groups 82 them and their methods, besides presenting challenges, limita-83 tions, and possible benefits for each of their definitions. In a 84 similar 2019 paper, Marques et al. [13] discussed situated visu-1 alization from a decision-making standpoint. Regarding its aid 2 towards decision support systems, they made a literature anal-3 ysis discussing the current areas of application, benefits, chal-4 lenges, and opportunities. Marques exposes how they found sit-5 uated visualization data in decision-making contexts to be more 6 rapidly and intuitively explored than the counterparts tested, al-7 lowing for earlier detection of flaws and higher work produc-8 tivity. Unlike both of these previous works, we present a novel 9 ASV application and perform original user tests to validate the 10 different visualization and interaction methods used. [14] . Center: Nutrition data situated on food products [15] . Right: GPS routing data on the actual streets [16] .

In recent years, multiple works tested the efficiency of data cluding gains in accuracy [14] , lower time taken [15] and lower 17 cognitive effort levels [16] . All of the tests involved simple day- 18 to-day tasks, such as picking a place to sit, buying groceries, 19 and following a GPS route. The present work, on the other 20 hand, puts ASV into an industry context, having tasks being 21 performed on real data, extracted from state-of-the-art commer- 22 cial equipment. Besides that, the multiple methods proposed 23 are tested against each other in an all ASV context, in order to 24 establish their effects on the task performance. 26 Only very recently, several works came in favor of further- 27 ing the exploration of freehand gestures in AR and VR contexts 28 [17, 18, 19] . Satriadi et al. [17] performed user studies compar-29 ing novel freehand interaction techniques in multiscale naviga-30 tion tasks, such as pan and zoom in a digital map. Their results 31 exposed a positive influence in user fatigue for their rate-based 32 input mapping technique, however with a trade-off for task 33 completion time. Kang et al. [18] compared object selection 34 and translation in three techniques. Interviews with the subjects 35 exposed that their direct touch and grab method provided them 36 with a higher sense of enjoyment and discoverability. Unlike In relation to other forms of interaction, two very recent 43 works make use of gazing in MR contexts [20, 19] . Lu et al. 44 [20] compared three different forms of accessing content using 45 eye and head movements. In their user tests, the eye-glance 46 technique was preferred by the subjects in long monitoring tasks, while also exposing the lowest results for the time taken 48 to acquire information. Meanwhile, Chen et al. [19] explored 49 the use of gazing movements as a means to select between dis-50 ambiguation options while the user's hands are already being 51 used. Their user study revealed the head movement to be the 52 overall preferred technique. Following their line of thought, 53 the current paper will test a similar gazing technique, however 54 aimed at a refined data-extraction context, testing its precision. 

Regarding large devices in their integrity, entire rooms may 58 be required for testing, especially when the interference be-59 tween multiple setups need to be analyzed. While conducting 60 this sort of experiment, total isolation between the test space 61 and the outside electromagnetic environment is recommended. 62 According to [2] , it is undesirable (and in some cases illegal) 63 to radiate high field strengths across whole bands of frequen-64 cies when conducting radiated susceptibility testing. In this 65 fashion, the use of screened chambers, as the one depicted in 66 Fig. 4 -right, became widespread. They are commonly built as 67 Faraday cages and lined with absorbing material inside, mak-68 ing them anechoic -i.e. rooms without any reflection of either 69 sound or electromagnetic waves. A large antenna, transmitters, 70 and receivers are used for the characterization of radiation pat-71 terns and EMC performance at different frequency bands. Ad-72 ditionally, circular platforms (turntables) are used to rotate the 73 devices during testing, as to capture a 360-degree view.

Having access to the measurements taken by one of these 75 fully equipped anechoic chambers, it is viable to render these 76 EMC readings in AR, situating it around the tested devices. 77 These readings are meant for users to detect any EMI that may 78 cause the equipment to malfunction or affect other systems. The 79 main advantage of such an application is to spatially expose ex-80 actly where the interference is being propagated from and into 81 what other components. In a situated view, it will be possible to 82 perceive the influence of multiple devices on each other in loco, 83 either inside the chamber during tests or anywhere else these 84 devices will be located at later on, by transporting the virtual 85 renderings along with the real components.

In a preliminary attempt to develop a prototype of the pro-87 posal, the physical room was scanned into a 3D mesh (Fig. 5) , 88 with its points in space being used as anchors for the data to 89 be placed upon. By loading this environment into an AR HMD 90 with spatial tracking capabilities, the mesh can be matched with 91 In a demonstration made to three experts in the EMC testing 5 area, the feedback was outright positive. Users commended the 6 visualization presented as being highly useful for analyzing real 7 data, even at a commercial level. This demonstration served as 8 a first evaluation of the concept. It also provided feedback from 9 the expert community, allowing for an understanding of their 10 needs, to be fulfilled in the upcoming steps. 12 Given the nature of the data measured by the aforementioned 13 EMC chamber analysis environment, two visualization patterns 14 1 https://docs.microsoft.com/en-us/hololens/hololens1-hardware were planned and implemented: 3D field topologies and 2D 15 color-coded ray casts. These were designed according to the 16 preliminary feedback given by EMC experts, who emphasized 17 the need to expose the scalability and reach of the field radi-18 ation. This is very relevant since understanding where (into 19 which components or devices) and how (with which intensity 2 20 or frequency) the radiation gets to is one of the primary tasks in 21 EMC testing.

3D field topology. As to supply the user with a broad view of 23 the field topology, a full three-dimensional mesh of the EMF 24 is produced based on the input data set. After converting the 25 spherical coordinate vectors into Cartesian points in space, this 26 object is rendered by drawing a line between every point and its 27 next neighbor. Each field presented in this view is read based 28 on a frequency given in the data set. Multiple frequencies of the 29 same field, or even different ones, can be compared by overlay-30 ing them, as can be seen in Fig. 6 , maintaining their topologies 31 and intensities relative to one another. Beyond that, it is also 32 possible to scale the fields by altering the logarithmic constant 33 used to convert the points in space from the spherical vectors. 34 This alteration slightly changes the topology of the field, while 35 also making it reach farther away from the center, properly cov-36 ering the points in space the actual EMF reaches. The loss in 37 intensity given the distance can also be calculated in this con-38 text. By doing this, the user can perceive which devices or 39 components the field hits, and with which intensity, allowing 40 them to fiddle with the environment setup and avoid undesired 41 EMI. The scaling interaction method implemented, as well as 42 other transform manipulation techniques, will not be further ad-43 dressed in this study as these did not take part in the user tests 44 performed.

2D Color-coded ray cast. Besides the full three-dimensional 46 fields, the EMC chamber analysis environment also provides 47 faster readings of planar sections of it. These are simply degree-48 by-degree measurements of the intensities in a plane at a partic-49 ular height of the EMF, in a given frequency. Since the goal is 50 to interpret where the EMI collides with different objects, these 51 vectors are rendered by drawing ray casts extending into infin-52 ity, as seen in Fig. 7 . The intensity of each vector is interpreted 53 both as a distance from the center (as to avoid occluding the 54 original tested device) and a color-scheme from least intense 55 to most intense. The decay in signal strength is also conveyed 56 in the loss of color intensity in each ray. They become more 57 transparent the farther away they are from the tested object. 

Given the primary task of analyzing the data to make assess-60 ments, having easy, precise, and uncluttered access to the values 61 in a data set is fundamental [21] . In order to narrow down an 62 optimal way for users to extract data from the visualizations, 63 three interaction methods were developed, to be later on tested 64 against each other in ASV-oriented data-extraction tests. Most 65 Hand Slider. Meant primarily for the color-coded ray cast vi-5 sualization, the Hand Slider method requires users to select a 6 single ray from the set, via a gaze and air-tap 4 combination, 7 3 https://www.nytimes.com/2020/05/28/well/live/whats-the-risk-ofcatching-coronavirus-from-a-surface.html 4 https://docs.microsoft.com/en-us/windows/mixed-reality/holograms-211 enabling a panel with an intensity measurement atop the line.

Users are prompted to slide one of their hands sideways, mov-9 ing the panel accordingly along the colored line selected (as in 10 Fig. 8) , exposing the different intensity values, much like in a 11 regular 2D UI slider. This is expected to provide a more refined 12 reading from the desired point in space. 13 Gazing. Similar to the previous method, the first step is to se-14 lect a single colored line from the visualization. Afterward, 15 whichever point of the line the user gazes at will display its in-16 tensity measurement. By moving their gaze along the line, users 17 will have access to the variations in its values, up to the moment 18 where they perform another air-tap, saving the last value read 19 and deselecting the line. Gazing is proposed as a quicker and 20 hands-free method, which can also be used from a distance, 21 as can be seen in Fig. 9 . This method is already widely and 22 commercially used for object selection in HMDs, having been 23 explored also in recent works [20, 19] . Touch Lens. Designed to be a more straight-forward metaphor, 25 the Touch Lens method acts as if the user's hand was a magni-26 fying glass. By simply laying their hands over or on a point in 27 space -virtually touching the data -users will have the read-28 ing from that point displayed on a hand-guided panel (seen in 29 Fig. 10 ). This was planned as being a more lifelike technique, 30 not demanding any complex or novel gestures from the user, 31 presenting a more localized view of the data. This method was 32 partly inspired by the work of Wagner Filho et al. [22] , where 33 a metaphor of touching virtual data in a scatter plot is used to 34 select it. 1 In order to assess the validity of the proposed solutions, two 2 user tests were designed and performed: a comparison between 3 three interaction methods and another of two visualization tech-4 niques. Each interaction and visualization method was eval-5 uated against the other methods in its respective group. In 6 both comparisons, users were asked to perform a series of data-7 extraction tasks in an ASV context. Their performances were 8 measured regarding their task correctness, time, and steps taken 9 during each trial. 10 Moreover, a series of questions were used to assess a set of 11 subjective aspects, such as the usability (using SUS [23] , the 12 System Usability Scale), the overall workload (using NASA 13 TLX [24] , the NASA Task Load Index), and any possible sim-14 ulator sickness symptoms (using SSQ [25] , the Simulator Sick- 

The experiment was designed to test the validity of the three 20 hypotheses described below.

alization methods on the self displacement users need to per- 23 form in order to accomplish data-extraction tasks in AR. 24 This hypothesis attempts to evaluate the distance users need 

For the tests, a prop antenna was used to situate real EMF 54 data around it, as shown in Fig. 11 -left. In the experiment 55 room, ten target objects were placed at different distances, ori-56 entations, and heights. One at a time, users were asked to mea-57 sure the intensity of the field when it touched one of these tar-58 gets, much like Fig. 11 -right. All of their measurements were 59 logged, as well as the time it took to complete the task and the 60 number of steps they took around the room, as a way to assess 61 their performances. Before and after every trial, subjects were 62 prompted to take a seat in a default position in the room. Apart 63 from that, users were free to walk and take as much time as 64 they wanted to complete each task, as well as use the interaction 65 method they were given whichever way they preferred. Prior to 66 the set of tests for each technique, subjects went through three 67 tutorial runs with it, as to allow them to learn how to operate 68 it. As previously mentioned, the tests were divided into two 69 comparisons, being separated in interaction and visualization 70 methods. The specifics for each of these are described below.

Interaction test. In this part of the tests, the interaction method 72 was used as an independent variable, maintaining the visualiza-73 tion fixed as the 2D colored ray casts. The three methods de-74 signed for data-extraction -Hand Slider (HaSl), Gazing (Gaze) 75 and Touch Lens (2DTL) -were tested against each other. This 76 test was designed to further develop the study of interaction 77 methods into an ASV context. By proposing and testing three 78 different techniques, the aim is to establish whether one of them 79 will have outstanding performance when compared to the oth-80 ers in each of the hypotheses proposed.

The order in which they were tested was alternated between 82 users, as to avoid a learning bias. For each user, three of the 83 ten targets were elected for each method, as to not repeat these 84 between conditions. Each of these three targets was repeated 85 three times to prevent outliers, using a Latin square to alternate 86 the order. For each technique tested, three of the neglected tar-87 gets were selected at random to be used in three tutorial runs 88 (one per target) right before each condition was tested, as to al-89 low the subject to learn how to operate it. After the nine trials 90 for each method were concluded, subjects were asked to answer 91 the aforementioned questionnaires before continuing to the next 92 condition.

Visualization test. Regarding the visualization part of the tests, 94 the visualization technique was used as the independent vari-95 able, fixing the interaction method as the Touch Lens. The two 96 visualization designs -3D field topology (3DTL) and 2D color-97 coded ray casts (2DTL) -were tested against each other. This 98 assessment was developed in order to further explore different 99 visualization techniques into the situated AR context. The ob-100 jective is to demonstrate whether one visualization presents a 101 significant increase in performance when compared to the other 102 for each of the hypotheses proposed, maintaining the same in-103 teraction method for both.

Again, the order of the methods was alternated between 105 users. Three of the targets were used for each method to prevent 106 repetitions between conditions, averting learning bias. Each of 107 the targets was repeated three times per condition in alternated 108 orders, avoiding outliers. In this test, the remaining unused tar-1 gets were used at random in three tutorial trials before each 2 method was tested, so the subjects could learn how to operate 3 them. After all trials in each condition, users answered the qual-4 itative questionnaires, before moving on to the next method. Walking Steps. As seen in Fig. 12 , a comparison of the number 38 of steps taken in the three different interaction conditions was 39 conducted. The same was done to the two different visualiza-40 tion conditions.

There was a significant difference in the number of steps 42 taken comparing the Hand Slider method (M=10.27, SD=4.69) 43 with the Gaze method (M=8.36, SD=5.78); t(44) = 2.08, 44 (p ttest = 0.041) and with the Touch Lens method (M=8.56, 45 SD=3.34); t(44) = 2.16, (p ttest = 0.033). This suggests a con-46 siderable decrease in the self displacement users need to per-47 form in order to accomplish the tests with the Gaze method 48 (18.61%) and with the Touch Lens method (16.66%).

Regarding the visualization methods exposed in Section 3.2, 50 although we observe a substantial increase in the mean number 51 of steps taken (23.63% from 2D to 3D), the results could not 52 demonstrate a significant effect (p ttest ≈ 1.3e − 1).

Time. In relation to the amount of time it took for the subjects 54 to finish the trials (Fig. 13 ), a significant difference was found 55 in a paired t-test when comparing the Hand Slider method 56 (M=36. 18, SD=31.25) with the Gaze method (M=20.93, 57 Once more, albeit exposing a substantial increase in time 8 taken means (87.6%), the comparison between the 2D and 3D 9 visualization methods was not statistically significant (p ttest ≈ 10 1.2e − 1). The lack of significance is probably due to the high 11 variance between subjects in terms of 3D familiarity combined 12 with the relatively small number of subjects. 13 Correctness. Based on the intensity measurements given by the 14 subjects during the trials, an error ratio is computed as a rela-15 tion between the user reading and the ground truth value. This In a paired t-test, a significant difference was found in the 

Usability. After the set of trials for each of the conditions 4 tested, a series of subjective questionnaires were applied to the 5 subjects. SUS [23] was used as a way to measure the usabil-6 ity of each method. In its score-based analysis 5 , the results are 7 available in Fig. 15 . In short, all techniques were deemed either 8 acceptable or marginal, with all three interaction methods being 9 rated above average in the 2D visualization. 10 Discomfort. The data collected for simulation sickness [25] re-11 vealed that there was not a high discomfort observed in any 12 of the tested conditions, as depicted in Fig. 16 . Three out 13 of the four conditions tested presented negligible symptom 14 results [26] . The Gazing method, although having reached 15 the level of significant symptoms (10.47 points, SD = 13.32), 16 scored far below the level of concern, since most of the subjects 17 reported minimal (20%) or no symptoms at all (40%). 18 Workload. With the objective of measuring different types of 19 effort exerted by the subjects, the NASA TLX test [24] was ap-20 plied. Considering the results exposed in Fig. 17 , it is notable 21 that all three interaction methods (using the 2D visualization) 22 were perceived as less demanding than the 3D visualization 23 technique. Besides, the Touch Lens interaction method in its 24 2D view was deemed as the least demanding in all workload 25 aspects. 26 Additional feedback. In a more general assessment, after each 27 test condition, users were asked to rate their agreement with 2. It was easy for me to navigate through the data. 4. It was easy to find the required information. 35 5. It was easy to remember how to do what I was asked. 36 6. Using the technique was comfortable. 37 When directly asked to rank the interaction methods tested 38 from most to least favorite, 60% of the subjects chose to place 39 the Gazing method in first, with the other 40% choosing Touch 40 Lens as their preferred interaction method. Regarding the visu-41 alization techniques, 60% said they preferred the 2D ray casts 42 over the 3D topology. In addition, two subjects complained 43 about the hand-tracking capabilities of the device, claiming it 44 lost track multiple times and that holding out their arm and 45 finger for the HMD to recognize them was annoying over the 46 course of the tests. 

Based on the results exposed, we make here a short analysis 49 of the possible meanings behind them. They are split below 50 between the Interaction and Visualization comparisons as a way 51 of discerning the contributions for these two fields.

Considering the interaction portion of the hypotheses pre-2 sented, all three were shown to achieve significant differences 3 in the results, partially proving all of them. In a short-sighted 4 view, it would be possible to rank the three interaction meth-5 ods according to each of the performance variables taken. This 6 would leave Hand Slider as the best technique regarding the 7 correctness, in the sense that it presented a significantly smaller 8 error rate than the others, in part confirming hypothesis H3. 9 Taking into account the manual refinement precision tool it of-10 fers, this is not a surprising result. It is relevant to note that it 11 also ranked as the most time-consuming among the interaction 12 techniques, which might indicate a precision-time trade-off. 13 In the same line of thought, Touch Lens ranked as the least 14 time-demanding method, with statistically significant results, 15 validating the interaction part of hypothesis H2. Since users 16 quickly realized that they only had to walk to the target and 17 place their hands there in order to perform a reading, their 18 cognitive load might have been diminished during these tri- 19 als, which might explain the decrease in time. This method 20 also presented favorable results for intensity correctness (with 21 the second smallest error rate), steps taken (second fewest aver- 22 age number of steps), and workload (smallest Task Load Index) 23 which are highly compelling results in its favor. Among the 24 techniques, the touching metaphor was also the most familiar 25 to the subjects, given its similarity to regular human behaviors, 26 which may explain the results. 27 The Gazing method came in first in requiring the least 28 amount of physical displacement by the user, supporting hy-29 pothesis H1. This makes a case for it being the overall most ad-30 vantageous way of interacting with data from a distance, espe-31 cially since it obtained both the best SUS score and overall user 32 preference among the interaction methods. The actual number 33 of steps taken during these trials is arguably due to data occlu-34 sion, either by physical objects or other pieces of data. This 35 could be mitigated with different techniques that minimize oc-36 clusion, such as making the data dynamically adaptive to the 37 user position [14] , or let the user interact with the occlusion, 38 such as filtering through walls or disabling it altogether [27] .

Although a fast and low demanding technique, gazing scored 40 very poorly in its correctness assessment, which suggests that it 41 is only viable for quick measurements, that do not prioritize ac-42 curacy. Another possible concern is how high its sickness score 43 was when compared to all others, which might indicate a slight 44 tendency to nausea from excessive head movement. 

It is important to note that adding a third dimension to a problem is expected to increase its difficulty. In relation to that, the 48 results expressed in the 2D ray casts and the 3D topology com- (Fig. 14) , on the other hand, a very significant effect of the 55 visualization method can be observed, supporting H3. 56 Regarding the lack of a third dimension in the ray casts view, 57 we believe that allowing the user to manually segment the 3D 58 topologies into 2D planar cross-sections would bring out the 59 best in both visualizations [28] . This would both legitimize the 60 use of the planar ray casts as a fully spatial visualization ap-61 proach and make the topology view to provide better usability 62 and require a lesser workload demand, besides allowing for the 63 use of the other interaction techniques. Despite the very low 64 usability score and very high Task Load Index, 40% of the sub-65 jects still claimed to prefer the 3D topology view over its 2D 66 counterpart. Given the correctness assessment, the 3D topology 67 view clearly showed its value surpassing its two-dimensional 68 contestant, ranking as the most precise visualization when be-69 ing analyzed with the same interaction technique (Touch Lens). 70

Despite having treated the data with a statistical analysis ac-72 cording to the number of samples collected in order to better 73 understand it, all results exposed in this work should only be 74 interpreted as anecdotal evidence. Given the current COVID-75 19 pandemic, the number of test participants had to be limited 76 to a small population size as to comply with the local health and 77 safety guidelines. Broader tests should be implemented in the 78 future.

Regarding the 3D visualization, non-expert subjects in our 80 study might have been misled into interpreting the topologies as 81 a volume, due to their virtual mesh. Electromagnetic fields are 82 continuously broadcasted in all directions, however with differ-83 ent intensities, which could also be showcased as vectors. This 84 sense is better expressed in the 2D ray casts view. The 3D topol-85 ogy into 2D planar field segmentation is arguably a reasonable 86 way to rectify this limitation.

As to properly evaluate the use of ASV in the EMC field, 88 a formal specialist user test is still required. The next step in 89 this sense would be to assess the proposed visualizations with 90 experts with EMC backgrounds, in a set of interference avoid-91 ance tasks. This test should be held inside an anechoic chamber, 92 where subjects will analyze the EMI data between two different 93 antennas and move them around to get an optimal placement, 94 minimizing interference.

This work presented the proposal, development and analysis 97 of ASV interaction, and visualization methods aimed at aiding 98 decision-making in an EMC testing context. The use case ap-99 plication is intended for helping expert users to analyze electro-100 magnetic fields and EMC data in general. Using an AR HMD, 101 users were able to visualize the spatial data read by high-level 102 industry standard EMC equipment in a series of task-based ses-103 sions, having their performances assessed in multiple ways.

The approaches presented in this paper demonstrated to have 105 different effects on data-extraction tasks. The least error-prone 106 and effort demanding (in time and user displacement) meth-107 ods were exposed, suggesting that specific techniques may be 108 used depending on the task priority. Suggestions have also been 109 made as to proceed with further design by combining different 110 methods and testing the application in real EMC assessments,

Introduction to Electromagnetic Compatibility (Wiley Series 28 in Microwave and Optical Engineering)

A handbook for EMC testing and measurement. Stevenage: 31 The Institution of Engineering and Technology

Situated visualization in augmented reality

Decision 36 making with visualizations: a cognitive framework across disciplines

The neural basis of choice and decision making

A time-dependent vector field topol-43 ogy based on streak surfaces

Physics-based visual characterization of molecular interac-47 tion forces

Visualization in meteorologya survey of techniques 51 and tools for data analysis tasks

Visualization of electromagnetic 54 field distribution with augmented reality

Measuring, logging, and visualizing pulsed electromagnetic fields com-59 bined with gps location information

Autonomous 63 electromagnetic mapping system in augmented reality

Situated visualization in the decision process through augmented reality

23rd International Conference Information Visualisation (IV

A usability assessment of augmented situated visu-75

2020 IEEE Conference on Virtual Reality and 3D User

Situated analytics: Demonstrating immersive analytical tools 81 with augmented reality

Development and usability analysis of a 86 mixed reality GPS navigation application for the microsoft HoloLens

Advances in Computer Graphics

Augmented reality map navigation with freehand gestures

IEEE Conference on Virtual Reality and 3D User Interfaces (VR)

A comparative analysis of 3d user in-94 teraction: How to move virtual objects in mixed reality

Conference on Virtual Reality and 3D User Interfaces (VR). 2020

Disambiguation techniques

for freehand object manipulations in virtual reality

ference on Virtual Reality and 3D User Interfaces (VR). 2020

uating information access methods for head-worn augmented reality

IEEE Conference on Virtual Reality and 3D User Interfaces (VR)

Data visualization: principles and practice

Virtualdesk: A comfortable 107 and efficient immersive information visualization approach

Sus: A quick and dirty usability scale

Development of nasa-tlx (task load index

Human Mental Workload; vol. 52 of Advances in Psychology

Simulator sick-116 ness questionnaire: An enhanced method for quantifying simulator sick

Configural scoring of simulator sickness, cybersickness and 121 space adaptation syndrome: Similarities and differences?

Situated visualization in aug-123

mented reality: Exploring information seeking strategies

Systems (SITIS)

Inferring cross-sections of 3d ob-127

jects: A 3d spatial ability test instrument for 3d volume segmentation

Proceedings of the ACM Symposium on Applied Perception