Interaction with a 3D modeling tool through a gestural interface: An Evaluation of Effectiveness and Quality

(1)

Thesis no: MECS-2014-13

Interaction with a 3D modeling tool through a gestural interface

An Evaluation of Effectiveness and Quality

David Gustavsson

Faculty of Computing

Blekinge Institute of Technology SE–371 79 Karlskrona, Sweden

(2)

weeks of full time studies.

Contact Information:

Author(s):

David Gustavsson

E-mail: dagb09@student.bth.se

University advisor:

Dr. Veronica Sundstedt

Department of Creative Technologies

Faculty of Computing Internet : www.bth.se

Blekinge Institute of Technology Phone : +46 455 38 50 00 SE–371 79 Karlskrona, Sweden Fax : +46 455 38 50 57

(3)

Abstract

Context. Gestural interfaces involves the ability of technology iden- tifying and recognizing human body language and then interpret this into commands. This is usually used to ease our everyday life, but also to increase usability in for example mobile phones.

Objectives. In this study the use of a gestural interface is evaluated as an interaction method to facilitate the introduction of new and novice users to 3D modeling tools. A gestural interface might reduce the modeling time without making an impact on the quality of the result.

Methods. A gestural interface is designed and implemented based on previous research regarding gestural interfaces. Time and quality results are gathered through an experiment where participants are to complete a set of tasks in the modeling tool Autodesk Maya that relates to modeling. These tasks are executed in both the implemented gestural interfaces as well as the standard interface of Autodesk Maya.

User experience is also gathered through the use of a SUS questionnaire.

Results. 17 participants took part in the experiment. Each participant generated time and quality results for each task of the experiment for each interface. For each interface the user experience was recorded through a SUS questionnaire.

Conclusions. The results showed that the use of a gestural interface did increase the modeling time for the users, indicating that the use of a gestural interface was not preferable as an interaction medium. The results did show that the time difference between the interfaces was reduced for each completed task, indicating that the gestural interface might have an increase in learnability of the software. The same indication were given from the SUS questionnaire. Also, the results did not show any impact on the quality.

Keywords: 3D modeling, gestural interface, gesture design, Leap Motion.

i

(4)

I like to thank my supervisor Dr. Veronica Sundstedt for her extensive work, great support and important feedback throughout the development of this thesis, without it should not have been possible to finalize.

Educator assistant Jonas Petersson’s help and insight in the Autodesk Maya API was of utmost importance in the development of the gestural interface.

The feedback from and discussion with Christopher Eliasson, Magnus Lindberg and Hans Lövgren have been a big help in the writing of the thesis.

I also want to thank my opponent Robin Lindh Nilsson whose feedback made important improvements on my thesis.

Marc Nicander did a wonderful job with the illustrations in the thesis, sacrificing his spare time to help me visualize my words.

Tobias Ljungkvist did a great job when helping me proofread my thesis, to give it the last push towards greatness.

Finally, my family are and have always been my main motivation for my hard work and they have always kept me going.

Thank you all.

(5)

List of Figures

1.1 The Autodesk Maya Interface. Copyright of Autodesk. . . 4

2.1 User Preferred Gesture Mapping . . . 13

3.1 The Leap Motion Device. Copyright of Leap Motion, Inc. . . 16

3.2 The Camera-Pivot Relation . . . 18

3.3 Supported Gestures. Copyright of Leap Motion, Inc. . . 20

3.4 Finger Visualization . . . 22

3.5 Camera Gestural Controls. . . 23

3.6 Autodesk Maya’s Gizmos. . . 24

3.7 The Interface Menu Structure . . . 26

3.8 Camera Translation Feedback. . . 30

3.9 Picking Controls . . . 30

4.1 The Tasks of the Experiment. . . 36

4.2 The Flow of the Experiment. . . 39

5.1 The Experiment Time Results. . . 44

5.2 The Camera Quality Results. . . 46

5.3 The Mesh Quality Results. . . 47

iii

(6)

4.1 Experiment Task List . . . 35

5.1 SUS Score Alternative Meaning . . . 43

5.2 P-value for the Time Results . . . 45

5.3 Percent Difference of the Time Result . . . 45

5.4 P-value for the Camera Results . . . 46

5.5 P-value for the Mesh Results . . . 47

5.6 Average of the SUS Test . . . 48

A.1 Summarized Pilot Study Results. . . 58

A.2 Summarized Time Results. . . 59

A.3 Summerized Quality Results. . . 59

A.4 Summarized SUS Results with Average, Usability and Learnability. 60 B.1 The Questions of the SUS Questionnaire. . . 72

iv

(7)

Introduction

By using and reading body language people can communicate to each other without the use of words. The same is possible between humans and computers, where gestures based on movement and placement of hands and body can be read and interpreted by different hardware devices and software implementations. To evaluate the effectiveness of modeling on new and novice users in regards of speed, a gesture based interface is evaluated when performing the task of 3D modeling.

The results give an insight into the impact on effectiveness when implementing a gestural interface into a 3D modeling software for new and novice users experience in such a tool.

Developers of 3D tools will be able to use this to decide wether to implement support for a gestural interface in their software, as that it may help introduce new users, or speed up the work for novice users, to the tool.

With the aim to ease the introduction of a 3D modeling tool this study will evaluate the implementation of a gestural interface in such a tool.

1.1 Topic Background

Here a presentation of the topcis involved in this thesis is given. The different knowledge areas applied in this thesis are the combination of 3D modeling tools and gestural interfaces. A presentation of the current use of 3D modeling and the modeling tools in the industry and society. A short review of gestural interfaces and the every day usage of such are then given as well as a briefing regarding different gesture reading devices.

1.1.1 3D Modeling Tools

Throughout the history of mankind, there has been an interest to express oneself with the help of artistic expressions. Through paintings, sculpting, stone masons and even great architectural wonders, people have had a desire to present to others their skills and creativity. During the last 200 years the concept of photographs, and later on movies, presented a new way to express one’s artistic skills. In more recent years, the concept of 3D art has established itself with the help of

1

(11)

Chapter 1. Introduction 2 games and movies. The use of 3D models in animated movies and games has grown into a vast industry with the success of animation studios like Pixar and game studios like Blizzard Entertainment. With this, the interest of expressing oneself through 3D model creation has increased. With well established tools like Autodesk Maya and 3D Studio Max, and free modeling tools like Blender, the interest and possibility to create models have increased and extended beyond the industry and into peoples‘ homes.

1.1.2 Gestural Interfaces

Gestural interfaces are a way to interact with computers through gesture recognition. Gesture recognition is the process of interpreting human gestures through mathematical algorithms. These are usually hand and face based. A gestural interface is when gestures are mapped against functionality in different devices and software, which enable communication through gestures; human-computer interaction (HCI) [19]. Gesture recognition is based on motion detection, where an input device, e.g. a camera, captures the surroundings, and an application tries to identify relative changes and interpret the input data as gestures. Primi- tive implementations of the technique is used in automatic door opening, traffic lights etc. More sophisticated examples are implementations of the motion controls in game consoles like Xbox through Kinect and Nintendo‘s Wii [9]. The software, the game in this context, has different predefined gesture input data that is compared against to be able to execute different commands in the game.

1.1.3 Gesture Reading Devices

Gestures are identified based on relative changes within a specific area. This area could be a road, a living room, or even a small space in front of a computer screen.

Changes within this area are recorded through different devices, and interpreted by software to be able to identify gestures. The changes can be registered by different kind of devices. Examples of different input data are image oriented as with a camera, distance changes recorded by infrared light or special devices like a glove, where the glove can describe the position, orientation and finger state of the hand [27].

Different devices have recently been developed with appurtenant SDK to enable implementation of gestural interfaces in existing and new applications. Ex- amples of these devices are Leap motion [3] and the upcoming Myo [5]. With the introduction of these there may be an upswing in the integration of gestures into the everyday life.

(12)

1.2 Motivation

To be able to interact with computers in a way that allows the computers to

"understand" humans have long been a sought after dream, or nightmare, to the public. The knowledge so far allows for doors to open as we approach them, lights to switch on when we move in dark areas, and calling friends by telling our phones their names. This way of communicating with technology and computers have eased the day-to-day life of humans, and could almost be considered a lazy luxury. But that does not limit the fact that our interest in facilitating our everyday life by allowing computers to "understand" us, and our needs. Almost every mobile device (phones, tablets etc.) uses some kind of gestural interface. With the focus on screen gestures on cell phones and tablets, also called pen computing, the intuitiveness and interaction have enabled even children to handle and use devices.

Taking advantage of the intuitiveness presented by the gestural interface, this study aims to evaluate the free hand gestural interfaces with regards on effectiveness and its impact on the resulting quality, with the hopes to ease the introduction to new and novice users in the area of 3D modeling. A more natural and intuitive interface may create a more inviting and available environment to an aspiring 3D artist. A favorable result of this study may encourage the integration of a gestural interface in current and upcoming 3D modeling tools as it may reach a larger audience of new users.

1.2.1 Problem Description and Statement

The process of modeling is a creative one, but the tools are often technically directed. The interfaces are usually vast, with a comprehensive set of functionality that can be daunting, and create a steep learning curve for a new or novice user.

Figure 1.1 displays an overview of the interface of the 3D modeling tool Autodesk Maya.

While the interface is needed to support the full functionality of the tool, the novice user only needs a small subset of the complete function list. These should not halt the user’s creativity by browsing menus, navigating through buttons and switching between controls. One approach is to utilize gestural interfaces to increase the intuitiveness and immersion for the user that the gesture interface offers [14][8].

1.2.2 Knowledge Gap

Work has been performed for developing and introducing gestural interaction within the process of modeling [14][8]. Evaluation of gestures has been made in regards of percieved performance, learning and fun [6]. The use of gestures made the user adapt quickly to the interface and the program. What is lacking is an

(13)

Chapter 1. Introduction 4

Figure 1.1: The Autodesk Maya Interface. Copyright of Autodesk.

evaluation of a gestural approach regarding effectiveness in regards of modeling speed. With the increased interest of gestural interfaces as well as the increased availability of gesture reading devices, the evaluation of such an interface versus the current method of mouse-keyboard is required. This to establish if gestural interface is a valid approach regarding the effectiveness of 3D modeling. As the industry has quality demands regarding the result, a potential increase in effectiveness should not affect the quality of the result.

This paper will evaluate the time and quality impact of implementing a gestural interface in a modeling tool. By using an industry standard modeling software as implementation platform, a connection to the industry will be created. The physical and economical aspects of the suggested approach also need to be con- siderd to make the suggested approach available to the target audience.

1.3 Research Question

The research questions of this thesis are the following:

• Does the use of a gestural interface reduce the creation time when working in a modeling tool for new and novice users?

• Does the use of a gestural interface affect the quality of the produced result?

(14)

Answering these questions will give an insight of the impact of implementing a gesture based interface in a modeling software.

1.3.1 Value Definitions

The two values mentioned in the research question above that are to be measured and evaluated are creation time and quality. The two values are given a short description here.

Creation Time

The definition of effectiveness in this regard is the time it takes to create a desired result. As the goal is to introduce new users to modeling tools, the area of interest is the effectiveness of modeling in such a tool. So the effectiveness is measured in creation time, the time it takes to model a desired object in a 3D modeling tool.

Quality

The quality of the results is a bit trickier. As 3D modeling can be categorized as art, the quality lies in the eye of the observer. To be able to measure the quality a defined result or goal has to be replicated and the amount of error towards the defined goal will define the quality. A perfect match (error value equals zero) is a desired result. A more detailed definition of quality is given in Chapter 5.

1.4 Objectives

To evaluate the gestural interface for modeling, such an interface is designed and implemented in a 3D modeling software. The functionality supported by the interface will be based on the process of box modeling. Box modeling includes, but does not limit to, transformation of objects and vertices, as well as extrusion of faces. The interface supports a subset of functionalities for box modeling, as all functionality is out of the scope of this study.

With the use of previous research and a pilot study, the gestural interface is designed around the selected functionality. This results in a final design, wich is the focus of the experiment, evaluating the gestural interface in regards of modeling speed. The experiment is a user case study, where the users have had no or very little previous knowledge regarding modeling and modeling tools. The participants perform a set of tasks for both the gestural interface and the standard interfaces in the modeling tool in focus. From the experiment the execution time from each interface is recorded, as well as the resulting scenes. These scenes are used to verify the quality from the different interfaces.

Finally a questionnaire evaluates both the implemented gestural interface and the standard Maya interface in regards of usability and learnability.

(15)

Chapter 1. Introduction 6

1.4.1 Target Modeling Software

The paper will use the software Autodesk Maya as the implementation platform to test the gestural interface on. Maya is a well established modeling tool within the industry of both movies and games. Used in the production of games like Halo 4 and Amnesia: The Dark Descent, and also for special effects in movies like The Girl with the Dragon Tattoo, Up and Frozen, makes Maya a good candidate to run the experiment on.

1.4.2 Target Reading Device

This study will use the Leap motion as input device to register gestures. The Leap motion can register 6 degrees of freedom, known as 6 DOF. Specifically, the Leap motion can register movement as forward/backward, up/down, left/right combined with rotation about three perpendicular axes; pitch, yaw, and roll, with addequit accuracy [25]. It also supports identification of fingers, hands and pointlabels, like pens and sticks. Finally, the pricing of the Leap motion as well as its small size makes it available for the target group.

1.5 Thesis Overview

Chapter 2: Background and Related work will give a view on the previous work relating to gestures and presenting different areas of use. Then work relating to gestural interfaces and its impact on the target research objects will be highlighted and discussed.

Chapter 3: Gesture interface is where the functionality supported by the interface is presented. The design process of the gestural interface is also studied with the different steps and decisions discussed and argued for. Those steps are the initial design, the pilot study and its results, and the final interface design.

The implementation structure will be presented, and difficulties and problems are listed and discussed here.

Chapter 4: Method discusses the experiment design, and the execution of the experiment are disclosed here. The data that is gathered are also listed and described in greater detail as well as the approach of its gathering.

Chapter 5: Result and Analysis presents and discusses the summarized results. The execution time, the resulting quality and the summarization of the SUS questionnaire will be presented here, as well as a discussion about the interpretation of the results. Possible problems that could have affected the results are also presented and discussed.

(16)

Chapter 6: Conclusion and Future Work presents what conclusions can be drawn from the results and analysis. Future work and improvements are also discussed.

(17)

Chapter 2 Background and Related work

Chapter 2 will provide background information about 3D modeling and the use of modeling tools, gestures and the use of gesture interfaces. Different approaches regarding gesture input devices, application areas and usage will be presented. In the later part of the chapter, related work to this thesis subject will be presented.

2.1 3D Modeling

3D modeling is the process of defining a mathematical representation of a shape in a virtual environment. This representation can later on be visualized in aspects such as games, movies or even architectural designs.

2.1.1 Primitives

The mathematical representation defining a shape can be categorized into different building blocks that makes up the shape. These building blocks are in its most atomic level, positions in a 3D space, described by their location in the con- taining space; x y z. These positions, referred to as vertices, can have different attributes beyond the positional one; color, UV coordinates, bone weights etc.

These attributes are not included in this thesis, as the focus is on the transformation, and therefore position, of the vertices. The vertices are defined as geometric objects, also named geometric primitives. Combining and connecting geometric primitives can create new geometric primitives. A connecting line between two vertices is considered an edge. An area stretched between three or more vertices is considered a face, and one or several connected faces is considered a mesh. These primitives are defined to ease the modeling process for a 3D artist. For example, by allowing a user to transform a face instead of a vertices, the artist can keep the geometric relations intact between the primitives (vertices), that defines the face.

8

(18)

2.1.2 Modeling Techniques

There are several different techniques when modeling. Some of them are listed here:

• Box Modeling is a polygonal modeling technique in which the artist starts with a geometric shape (cube, sphere, cylinder etc.) and then refines it until the desired appearance is achieved.

• Edge/Contour Modeling is another polygonal technique. In edge modeling the model is essentially built piece by piece by placing loops of polygonal faces along prominent contours, and then filling any gaps between them.

• NURBS/Spline Modeling creates meshes that are comprised of smoothly interpreted surfaces, created by "lofting" a mesh between two or more Bezier curves (also known as splines).

• Digital Sculpting creates meshes organically, using a tablet device to mold and shape the model almost exactly like a sculptor would use rake brushes on a real chunk of clay.

• Procedural Modeling creates scenes or objects based on user definable rules or parameters.

• Image Based Modeling is a process by which transformable 3D objects are algorithmically derived from a set of static two-dimensional images.

• 3D Scanning is a method of digitizing real world objects when an incred- ibly high level of photo-realism is required.

3D modeling is an extensive topic and an in depth description of different modeling techniques is out of the scope of this article. The focus in this thesis will be the technique of box modeling. Box modeling is a modeling method that is considered a relative quick and easy modeling technique to learn.

Box Modeling

A 3D modeling technique in which the artist begins with a low-resolution shape and modifies it by extruding, scaling, or rotating different types of primitives.

Extruding faces and sub-dividing faces are some of the techniques that relates to this type of modeling [1].

Extrude faces: The extrude face functionality encapsulate the process of

"dragging" a face from its position, while new faces are generated to keep the face connected to the original geometry. This allows the user to extend the geometry, generating more vertices and faces to better define the surface of the model.

(19)

Chapter 2. Background and Related work 10 Subdivision: Subdividing a face or an edge splits the affected primitive into smaller primitives by adding new vertices. This allows the user to increase the detail of the surface.

2.1.3 Gizmos

Gizmos are not part of the ’physical’ scene or primitives, but acts as a representation of an operation, and gives the user information about the status of that operation. This could be a translation gizmo, that allows the user to see what direction a translation operation acts. As declared by Kim et al. [14], gizmos generate robustness to the interaction, as well as feedback of the active operation. The gizmo is controlled by the user, and the primitives are controlled by the gizmo.

2.1.4 Virtual Environment

A virtual environment is defined by Schroeder as:

“a computer generated display that allows or compels the user (or users) to have a sense of being present in an environment other than the one they are actually in, and to interact with that environment” [22]

That fits well with a 3D modeling tool where the user tries to express different objects within the scene of the modeling tool. When interacting with virtual environments, there are three universal tasks that are the focus of interaction [7]:

• Navigation

• Selection

• Manipulation

Modeling is no different. The user has to be able to navigate in the scene to find the primitives of interest. These primitives have to be defined as the target of manipulation. This is done through the selection of said primitives. While selected, manipulation of the selected primitives is the process of modeling. These different tasks need to be supported to be able to evaluate the gestural interface in the virtual environment.

2.2 Gestural Recognition

Gestural recognition is a topic within the field of computer science with the goal of interpreting human gestures via mathematical algorithms. This development of HCI has several different areas of use. It can be used as an accommodation tool through the use as sign language recognition, for entertainment as interfaces

(20)

in games or as virtual controllers to increase effectiveness and / or availability.

This section will present a subset of different reading devices that are used when gathering gestural data. Then a small overview of different interpretation algorithms are presented, as well as different gesture types. Gestural interfaces are described shortly, as well as different areas of use. Known problems relating to gestural interfaces are discussed in the end.

2.2.1 Reading Devices

The process of reading the gesture data involves gathering the state of a volume of space, whether it is a car lane, a living room or the space in front the computer screen. The position of the space is not always static, as the space can be relative the reading device. The type of data gathered can also vary. Pictures and videos are one type of data (image data) that can be interpreted. Objects position and rotation, another type of data, can be gathered through magnetics or inertial tracking devices. This data is gathered by different reading devices. Some are listed here:

• A wired glove is a device that is worn on the hand of the user. The glove is able to track position, rotation, and in some cases, even read the degree of bending of the fingers [27].

• Cameras are used to gather image data. There are also depth-aware cameras that are able to capture the depth state of the space. The Xbox Kinect are a combination of a regular camera and a depth camera.

• When combining two or more cameras, stereo-cameras, it is possible to create a 3D representation of the space.

• Controller-based gestures are based on devices that act as an extension of the body. The computer mouse is an example of this type of gesture device.

Other examples are the Nintendo Wii Remote and Myo [5]. An early type of controller was the Polhemus Isotrack [10].

Some reading devices are released to the public with a complementary software development kit (SDK) to allow developers to integrate gestures into their application. Xbox Kinect released an SDK in June 2011, allowing developers to integrate Kinect into own created applications. Leap motion [3] and Myo are other examples of devices released with an SDK.

2.2.2 Reading Algorithms

There are different types of algorithms that are able to interpret input data as gestures. Pavlovic [19] defined two different system models when interpreting the hands; 3D model-based [11] and appearance-based [15]. The first method makes

(21)

Chapter 2. Background and Related work 12 use of 3D information of the body parts in order to obtain several important parameters, like palm position or joint angles. Appearance-based systems use images or videos for direct interpretation. While the 3D information can retrieve more extensive detail and information, it is more computational heavy.

2.2.3 Gesture Types

After reading and interpreting the data this information is available to be mapped to functionality in the interface. Kammer et al. [13] defined two different types of gestures that are to be considered; online gestures and offline gestures.

Online gestures are input that are directly mapped to the interface. This can for example be translation or scaling operations.

Offline gestures are not processed until finished. This is what is related to actual gestures. A circular motion is not noticed by the interface until the circle is complete.

2.2.4 Gestural Interfaces

Gestural interfaces are a set of gestures mapped to functionality in a software.

This allows the user to interact with software through the use of hand and body language. There are different areas of usage. Every day use of gestural interfaces is through motion detection, allowing doors to open when approached, lights to switch on when detecting motion and traffic lights to switch only when there are cars nearby, improving the traffic flow. Another common application is the use of gestures in mobile phones and handhelds to simplify the use of these.

Gestural interfaces are also used in an entertaining manner. The game consoles of today are usually shipped with an integrated gestural reading device. One of the early releases of gestural devices was the Nintendo Power Glove. While it presented an alternative way of control, only a few games were released with gestural interfaces in mind. This made controlling the games with the device unintuitive and difficult. Not until 2006, with the release of the Nintendo Wii, had the gestural interfaces made a breakthrough into the homes of the public.

Here the design of the games was centered around the gestural input connected to the Wii, which made the controls feel more fluent and intuitive. This resulted in an immersiveness that appealed to the public [26]. Microsoft followed up Nintendo’s Wii with the release of Kinect to the Xbox.

There are several different areas of use, and far from all are mentioned here.

In the related work section the relation between gestural interfaces and modeling will be observed in greater detail.

(22)

2.3 Related Work

The related work section will first review the designing directives for gestural interfaces. The later part discusses work surrounding gestures in modeling software.

2.3.1 Gesture Design

While the gesture interface is a technological approach to developing a way of interacting, the main focus has to be on the users the interface is intended for.

Beurden et al [6] present an evaluation of gestural interaction where they do a user case study to evaluate how people choose to interact with software through gestures. She recorded what gestures a user preferred to complete different tasks.

The results are visible in Fig 2.1.

Figure 2.1: Overview over gestures most frequently used for various manipulations.

Reprinted from [6]

Based on this, Beurden developed an interaction technology for medical images. Beurden evaluated this based on a 7-point scale with word-pairs representing extreme ooposites, good to bad and easy to hard.

Pol et al. [24] performed a user case study exploring different methods to interact with virtual environments in regards of navigation, selection and manipulation. Regarding navigation, he found that head tracking was an intuitive and inobstructive way of navigating. For distance selection (not reachable within

(23)

Chapter 2. Background and Related work 14 arms length) ray picking was a good way to interact with the enviorment. Ma- nipulation should be done close to the body. Distant interaction was useful for rough manipulation.

Nielsen et al. declared that there is a need to incorporate the intended users when designing gestural interfaces. As stated by Nielsen:

“Gesture research shows that there is no such thing as a universal gesture vocabulary, so a good gesture vocabulary may only match one specific application and user group.” [18]

there is a need for the target group to be a part of the design to be able to develop an intuitive and complementary interface.

2.3.2 Gestures for Modeling

There has been a great interest regarding alternative interaction models in regards of modeling. A great focus has been around the process of sculpting, to be able to mimic the process of traditional clay modeling [10][12]. The arguments for this research are that this would increase the immersiveness and intuitiveness of sculpting, allowing users to use and learn the system at a quicker pace. To further increase the intuitiveness of the sculpting process, Sheng et al. [23] presented sculpting through proxy, where he developed a clay proxy that allowed the user to get physical feedback when interacting with the sculpting software. These papers focused more on the input devices rather than the development of the interface.

Dave et al. [8] presented a gestural interface for CAD where he mapped whole body gestures captured by the Kinect to manipulate objects in a sculpting fashion.

He declared that the process of modeling was a two-step process, where the first step was concept visualization and the second was the final design. He stated that the standard approach of the first step was performed with clay modeling, and declared the need of a gestural interface for this step. He discovered that new users adapted to the gestural interface quickly.

In 2005, Kim et al. [14] developed a gesture interface to select, create, manipulate and deform mesh objects. With the tracking of head and fingertips with the help of markers, the user was able to manipulate mesh objects on a vertex level.

The system also allowed the user to 2D paint the intended object with the use of finger movement, and then the system would recreate a 3D representation of the painted result. They found that gestures alone does not make up for a complete system. The interface need to work in conjunction with different components:

• hand gestures

• gizmos for virtual object controls

• textual menus

(24)

Kim did not evaluate the implemented gestural interface in regards of speed and quality. While he observed that the gestural interface made it easy to introduce users to the interface and the supported functionality, it was not in relation to other interfaces.

(25)

Chapter 3 Interface Functionality and Design

Chapter 3 gives an overview of the implemented gestural interface. The first part is a walkthrough of the functionality supported by the gestural interface. The second part discusses the design of the gestural interface. This involves a first design based on previous research. The first design is then put through a pilot study that evaluates the initial design, and results in a final design implementation.

The chapter also discusses problems related to the interface design process.

3.1 The Leap Motion

The hardware used as gesture reading device is the Leap Motion. The Leap Motion is a small (1.3 cm x 3.1 cm x 7.6 cm) device that easily fits infront of any computer, and as the connection goes through a USB cable, the device is accessible to the majority of todays computers. Without any reading area specifications, it is claimed to be up to 8 cubic feet [3]. This allows for the reading of almost the entire area (2 * 2 * 2 feet) in front of the computer.

Weichert et al. [25] did research regarding the Leap Motion´s accuracy and found it to have an accuracy of 1.2 mm in a dynamic setup. Based on this, the Leap Motion is a good representation for a reading device in this study. The Leap Motion is displayed in Figure 3.1.

Figure 3.1: The Leap Motion reading device. Image property of the Leap Motion, Inc, who owns the copyright.

16

(26)

3.2 Supported Functionality

Implementing support for all the functionality involved in box modeling is out of the scope of this paper. The main focus in regards of box modeling will be the support of different types of primitives and the extrude face functionality. It should also cover the areas stated by Bowman [7]. Beyond these areas, menu navigation is also supported within the interface as stated by Kim et al. [14].

These areas can be considered states within the interface that from here on will be refered to as gestural states. This chapter will describe the functionality supported in the gestural interface and why it was supported.

3.2.1 Different Primitives

As stated before, meshes are a collection of several combined primitives. By changing the type of primitives the user chooses to work with, he/she can gain better control over the modeling process. The primitives supported in the interface are limited to vertices, faces and objects/meshes.

3.2.2 Selection

To be able to work with the primitives, the artist needs to be able to indicate what primitives he/she wants to work with. This is done by selecting primitives within the 3D scene. The artist can select any of the supported primitives, but is limited to one primitive type at the time. The interface also supports multi select, to be able to interact with several primitives at the same time.

3.2.3 Navigation

To be able to move around in the 3D environment within the modeling tool, the user has to be able to transform the camera that displays the objects within the scene. This functionality does include translation, zooming, rotation and refocusing of the camera on different objects.

Usually when modeling, only subparts of the entire scene is the focus of the user when modeling. These parts can be a specific model, specific faces or even specific vertices. The camera has to be able to focus on any of these subparts.

This is done by the use of pivot points. With transformations affecting a pivot point, this enables the user to focus the view on the average positions of the selected primitives.

All transformations of the camera relates to the pivot point of the camera.

See Figure 3.2. The translation, movement, of the camera affects the pivot point, not the camera. As the pivot point is translated, the relative position of the camera towards the pivot point is kept intact, indirectly moving the camera.

The translation acts in the transformation base of the camera. Sideways and

(27)

Chapter 3. Interface Functionality and Design 18

Figure 3.2: The camera movement in relation to the pivot point.

up and down relates to the camera´s current rotation. Zooming is also in the transformation base of the camera, but acts between the camera’s position and the pivot point of the camera. This hinders the camera to traverse behind the pivot point. When the camera approaches the pivot point the zoom speed decelerate down to zero. This keeps the camera´s focus on the pivot point.

Rotation, as translation, is focused around a pivot point. The rotation does not rotate the camera about its own axis, but rotates the pivot point the camera focuses on. This ensures that the camera never loses its focus when rotated. This further support the aim to let the user set the focus of the camera towards the subpart in the scene the user is currently working with / has selected.

The interface also enables the user to set and change the pivot point of the camera. When a subpart of the scene is selected, the user is able to set the selection as the focus of the camera. This centers the view of the camera to contain all of the selection. The position of the new pivot point is the average position of the current selection.

Lastly the interface also includes working with different views. Except for the perspective view, that enables the user to experience the scene in a 3D manner, there is also support for three different orthographical views; front, side and top.

With this the user can view the relation between objects in regards of position.

3.2.4 Manipulation

The process of 3D box modeling is based on transforming the different primitives in the scene. This is supported within the interface, through the functionality of translation, rotation and scale. This functionality is applicable to all the sup-

(28)

ported primitives.

Extrude Face

By supporting extrude faces the user can increase the details in the target mesh.

As box modeling is the process of refining a simple shape to more complex forms, there is a need for the user to be able to extend the geometry, generating more vertices and faces to better define the surface of the model. Extruding faces allows the user to do this [1].

3.2.5 Menu

Menu navigation does not relate directly to box modeling, but is needed to support the functionality of the interface. As stated by Kim et al. [14], textual menus are needed to be able to extract the full power of gestural interfaces. To be able to support several different functionalities, there is a need to navigate between the different gestural states mentioned above. The state transition will be supported through menu selection.

3.2.6 Undo and Redo

Based on the heuristic ’User control and freedom’ stated by Nielsen [17], the support of undo and redo is implemented, to allow the user to recover from unintended actions and errors that occur. As unintended actions are extra plausible in a gesture interface because of the possibility of executing unintended gestures, this support can be critical.

3.3 Interface Design

First an introduction to the supported gestures is given, as well as a description about their execution. Then the design is presented. The design of the interface was a three step process. First an initial design was implemented, based on the previous research and papers. The first implementation was then evaluated through a pilot study, where the implementation was evaluated based on execution time and observations. Based on the results, feedback and comments from the pilot study, the gesture interface was redesigned. The final design was then used in the experiment.

3.3.1 Gestures

The gestures used are mainly based on the gestures supported by the Leap Motion SDK. The gestures supported are both online and offline.

(29)

Chapter 3. Interface Functionality and Design 20 Online Gestures

The positions of hands and fingers are used troughout the interface to determine the relations between the hands and fingers and the interaction with the modeling tool. It is also used for manipulation of primitives.

Offline Gestures

Offline gestures are used when interacting with the interface. These gestures are identified by the Leap Motion and the following gestures are used throughout the interface; screen tap, key tap, circular gesture and swipe. The following descriptions are taken from the Leap Motion API [2]:

Figure 3.3: The gestures supported by the Leap Motion SDK. From left to right: screen tap, key tap, circular gesture, swipe gesture. Images are the property of the Leap

Motion, Inc, who owns the copyright.

Screen tap: A screen tap gesture is recognized when the tip of a finger pokes forward and then springs back to approximately the original postion, as if tapping a vertical screen. The tapping finger must pause briefly before beginning the tap.

Key tap: A key tap gesture is recognized when the tip of a finger rotates down toward the palm and then springs back to approximately the original postion, as if tapping. The tapping finger must pause briefly before beginning the tap.

Circular tap: A circle movement is recognized when the tip of a finger draws a circle within the Leap Motion device field of view.

Swipe tap: The swipe gesture is represented by a swiping motion of hand, a finger or a tool.

(30)

3.3.2 First Design

The first design was mainly based on three different papers. Beurden’s research was used to define what gestures were to be used in the interface [6]. The results from that study are visible in Figure 2.1. Pol evaluated different types of interaction method in a gestural interface [24]. Lastly, Kim [14] developed a framework for gestural interface when interacting with 3D environments. Kim recognized that to gain full power of gestures it had to be in conjunction with control gizmos and textual menus.

Screen Mapping

To be able to let the user navigate within the interface, and visualize what the gesture reading device (the Leap Motion) sees, the fingers and hands are mapped on the screen. This is done by projecting the relative 3D positions of each visible finger to the screen space of the modeling tool. This allows the user to register how many fingers is observed and where they are in relation to the reading device, i.e. where they are pointing. Each finger is equipped with a small point texture to visualize the screen position. While there is no exact measure of the volume where the Leap Motion register movement, tests found that following mapping was adequate:

screenX = (worldX/400 + 0.5) ∗ viewportW idth screenY = (worldY /400 − 0.25) ∗ viewportHeight

worldX as well as worldY are both measurements in mm and relates to the relative position towards the Leap Motion device. Figure 3.4 visualize the representation of a finger screen position.

Reusing Gestures

Kim evaluated a gestural interface by using only a few well-known hand gestures.

With a vast list of functionality, there is a need to be able to reuse gestures. For example pointing in selection mode should select or deselect primitives in the scene. But pointing when in translation mode should keep the selection intact, and allow the user to move the current selection around. And selecting in a menu should not affect the environment behind the buttons. As can be viewed by Figure 2.1, Selection and Activation share the same gesture. To be able to support this, two different approaches were implemented; states and finger count.

(31)

Figure 3.4: Top: The visualization of a registered finger within Maya. Bottom: Two fingers registerd in a Maya scene.

States: Different states involve setting the interface in different gestural states, wherein the gestures are mapped differently. By changing states the user can feel confident that the current state ensures that changes done in other states are intact, and will not be affected while working in the current state. States defined in this interface are selection, transformation and menu.

Finger count: The number of fingers visible dictates what different gestures mean. With this there is no need to change state, do specific gestures or enter the menu, and this generates a flow in the working process. The following mapping was done:

• To interact with the scene, one or two fingers should be present.

• Four fingers enable the user to cancel or undo / redo actions.

• When five fingers are present, the camera controls are active.

This allows the user to actively decide when he wants to influence the 3D environment, undo an error or rotate the camera.

(32)

Navigation

It was stated by both Kim and Pol that head tracking was the more intuitive way of navigation. This was not applicable in the interface of this paper, as there was no functionality for tracking the head. Beurden declares that zooming is preferably done by grabbing objects and dragging them towards you or pushing them away [6]. While this is an intuitive way that is executable with the hands, the support for navigating sideways, up and down as well as rotating the view is of interest.

Translation: By extending the gestures of zooming declared by Beurden, the relative position of the hands to the Leap Motion acts as a direction indicator for the camera. The direction in relation to the middle of the screen act as the direction of acceleration. Within the reading area of the Leap Motion device, a stop zone is declared. While in the zone, the translation of the camera decelerate until zero.

Rotation: Rotation of the camera is based on the normal of the hands palm.

Depending on the direction of palm normal in relation to the Leap Motion device, the hand is rotated around the pivot point.

Figure 3.5 illustrates the hand movement and orientation used to control the camera.

Figure 3.5: This illustrates the camera controls for the camera movement.

(33)

Chapter 3. Interface Functionality and Design 24 Picking

Picking is implemented to be able to interact within the interface. The picking is mainly done for selecting primitives, interacting in the menu and to activate and deactivate the transformation gizmos. The execution for picking is a combination of key tapping and screen tapping. To ’click’, a key tap is done to execute the activation. This executes the menu button activation, picking selection or activation of transformation gizmos. A screen tap activates a rectangular screen selection to enable multi select. Another screen tap finalizes the rectangular screen select.

Transformation

All transformation is executed through gizmos. Based on Kim’s statement, gizmos increases the robustness of the transformation operation. When entering any transformation mode, the transformation state becomes active. When in transformation states, the screen position of the hand dictates the position of the picking point and thus is able to interact with the gizmos. When over a gizmo, picking gains control of the hovered gizmo, and indirectly, the selected primitives. To deactivating the gizmo control, another picking action releases the gizmo. Each transformation operation is represented; translation, rotation and scale.

Figure 3.6: These are the transformation gizmos of Autodesk Maya. From left-to-right;

translation, rotation, scale.

Translation gizmo: The translation gizmo is represented by 3 axis aligned lines, which are connected to handles in the shape of arrowheads, where each line represents movement in the visualized axis. It also contains a quad in the center that enables translation in a plane aligned with the camera view. This gives the user better control of the movement, as it generates a reference point to the translation. When any of the arrowheads are used to move the object, the translation is done in the axis represented by the arrow.

(34)

Rotation gizmo: The rotation of the object is executed through the rotation gizmo. The rotation gizmo consists of four different circles, where three circles represent rotation around the local axis’ of the current selection, and the fourth is rotation around the view axis of the camera. The axis’ circle allows for the user to perceive the effect of the rotating process as it is executed, as he/she will see the rotation of the local axis’ of the selection while rotating. While the circles allow the user to rotate around a specific axis, there is also an invisible spherical volume that allow the user to freely rotate the object based on the axis difference between the initial piking position and the current position of the cursor.

Scale gizmo: The scale gizmo resembles in some ways the translation gizmo.

The handles of the scale gizmo are represented by cubes that are connected with lines that visualize the local transformation axis of the current selection. Scaling in any of the cubes results in scaling in direction of that axis. Here is also a center cube that represent uniform scaling (equal scale in all directions).

Each gizmo is visualized in Figure 3.6. The gizmos are equal to those within the Maya API, as the differences between the interfaces should be the interaction of the modeling program, not the behavior. As precision when picking is an issue when using gestural interfaces, the gizmos handles are scaled in the gestural interface to counter the precision problem. The scale is about five times the original size.

Menu

To support the transitions between states and different functionality, a menu system was implemented that allows the user to navigate in the interface. The structure of the menu is displayed in Figure 3.7. While working with the modeling tool, the menu is hidden from view. The menu is opened with a circular gesture of the finger and / or hand. The circular gesture is chosen to hinder ac- cidental opening of the menu while working with the interface. The whole menu is button based. That means that the interaction within the menu is through picking different buttons available. The layout of the menu is based on a circular placement around the center of the screen. The placement is evenly distributed based on the number of elements in the menu. This hinder the user accidently selecting unintended buttons. To close the menu, a four finger swipe closes the menu without activating any buttons. This in accordance with Deactivation in Figure 2.1.

Activating buttons: To indicate what buttons are to be selected, a line is drawn between the center of the menu (the center of the screen) and the screen position of the hand. The button intersecting the line is the one activated when

(35)

Figure 3.7: The structure of the menu implemented in the interface.

executing a screen tap. This counter the precision problem that occurs when interacting with gestural interface.

Other Functional Design

The interface also support extrude faces, focus camera, and changing the view between perspective view and three different orthographical views. These are all accessible through the menu and are represented by individual buttons.

Activating the corresponding button will execute the functionality described below.

Extrude face: The faces that are selected are extruded and the extruded faces becomes selected. Then the translation tool is enabled to allow the user to move the newly extruded faces. Movement is done through the translation gizmos.

Focus camera: When selected in the menu, the active camera is translated to focus on the active selection in the Maya scene. If no selection is active, the camera focuses on the origo of the scene; the 3D position (0, 0, 0).

(36)

Change view: Each view is represented by a button, and upon activation the current scene camera is replaced by the corresponding camera to what button is activated; front, side, top.

Lastly, the interface supports undo and redo. Figure 2.1 declares Deactivating as a manipulation action. This will be interpreted as a desire to cancel the current action, or regret what was previously done. By wiping in either left or right direction, the user is able to undo and redo, in respective order. As stated above, to enable either undo or redo, the Leap Motion has to see four fingers.

This hinders the user to accidently execute either undo or redo action.

3.3.3 Pilot Study

The first design presents a gestural interface based on research on intuitive gestures and directives from previous implemented gestural interfaces in modeling tools. But Nacenta et al. [16] state that user-defined gestures are usually easier to remember. As mentioned in Chapter 2, Nielsen declared that there is a need to incorporate the intended users when designing a gestural interfaces. This is attempted through a pilot study with participants from the targeted group. The purpose of the pilot study is split in two. First to evaluate and if needed, redesign the gestural interface. The second reason is to evaluate the user case study in terms of execution time. By executing the experiment based on the experiment design, an understanding of execution process and the actual execution time is created.

Structure

The participants of the pilot study are presented to the layout of the intended experiment. Information about their previous modeling experience is noted by the investigator. The participants are informed of the five different tasks that are the focus of the pilot study and will be used to evaluate the gestural interface.

These tasks are to be executed once with the gestural interface, and once with the standard Maya interface. Control schemes for both the gestural interface and the Maya interface are also supplied by the investigator on scene. Upon start of each given task with either interface, the time is recoded by the investigator. A total time limit on the pilot study is set to 60 minutes. This is due to the unknown total execution time of the tasks of the pilot study. The focus of the pilot study is not to let every participant complete all tasks, but to evaluate what is plausible to do during that time. The study was concluded when either the participant completed all the tasks, or the time limit was reached. It should be noted that none of the participants did complete all of the intended tasks in the pilost study, so different participants were given different tasks to be able to evaluate all of the five tasks. What participant that was given which task was determined by the investigator at the scene.

(37)

Chapter 3. Interface Functionality and Design 28 Execution

The participants for the pilot study were selected based on previous computer knowledge. This was ensured by selecting participants that did study, or had studied, software engineering at BTH. Participants both with and without modeling experience were selected. There were two participants with previous modeling experience, one with little modeling experience and four inexperienced. Each participant performed the study in an isolated office room with an investigator present. No information about the content or the structure of the study was presented beforehand. The participants were encouraged to ask questions, give feedback and comment on the experience throughout the study, to allow the investigator to record the session in as great detail as possible. Comments and observations were noted, and then summarized. The results are presented in Ap- pendix A. As stated previous, no one did complete all of the intended tasks, and therefore there are different tasks are timed for different participants.

Result

Based on the results of the pilot study, problems and possible improvements were observed. A textual summarization is presented below.

Finger occlusions: While moving and rotating the hand above the Leap Mo- tion controler fingers occasionally were occluded, and lost by the Leap Motion controler. As stated before, the number of fingers seen by the Leap controller dic- tated what action different gestures involved. This generated unintended actions.

For example camera transformations could halt, or picking stopped working.

Camera controls: The camera controls were perceived as too intensive, and without feedback on the relative position of the hand controlling the camera, the users had a difficult time controlling the camera. The participants never knew when the camera was about to move and when the camera was in a still state.

Using deceleration to stop the camera was also hard to control, as there was no indication on when the deceleration occurred.

Picking: The key tapping gesture used when picking generated a big problem.

While the gesture was intuitive and users were quick to learn the functionality of the gesture, the tapping of the finger made the Leap Motion vary the position of the hand during the gesture execution. This despite the participant holding the hand completely still. This made picking difficult as precision already was a problem. When the intended picking was executed, the position of the picking varied during the gesture, making picking near impossible. A more precision friendly approach was needed to allow the users to pick at the intended position.

(38)

The same problem was generated with the screen tapping gesture. To be able to screen tap, the user had to move the hand towards and back from the screen in a tapping motion. This proved to be hard to execute while focusing on specific positions. While the screen tapping was executed, the screen position representing the tapping position moved as the hand tried to execute the gesture.

Menu interaction: One observation was made where the interaction time in the menus could be improved. While not noted by any of the participants of the study, the interaction time in the menu was a bit lingering while searching for different buttons. As different menu levels exists, there was a bit of delay when transitioning between these levels.

Redesign of the experiment: The changes to the experiment design are highlighted in Chapter 4.

3.3.4 Final Design

Based on the pilot study the interface was redesigned. The focus for the redesign was the camera controls, the selection / picking gesture and the menu layout.

Camera Controls

To increase the control of the camera feedback was needed to allow the user to see when the camera transformation was active, when transformation was executed and show the direction in which the camera was traveling. There was also a need to decrease the acceleration rate of the camera translation to give the user time to react upon unintended transformation of the camera. To further increase the user’s control of the camera, the deceleration used to stop the camera translation had to be replaced with instant stop when the hand entered the transformation free zone.

Transformation feedback: When entering the camera transformation state through the revelation of five fingers to the Leap Motion, the modeling tool displays a screen rendered sphere at the bottom left of the screen. See Figure 3.8. When displayed, the user is informed that the camera is in transformation mode. Within the sphere a cone is drawn, indicating the hand´s position relative to the Leap Motion device. The sphere represents the transformation free zone, and upon moving the hand outside of this zone, the sphere is drawn in red. This indicates that the camera is currently accelerating in the direction in which the cone is pointing. This generates the feedback of when the camera is transforming,

(39)

Figure 3.8: This is the feedback sphere that informs the user when and where the camera translates. Left: Camera transformation is active, but not moving. Center:

Camera is moving right. Right: Camera is moving up.

and in what direction. This informs the userwhere the hand is, and where to move the hand to stop the transformation.

Picking Gesture

To be able to pick with precision, gestures not involving quick hand movements were needed. In the meantime the user need feedback when the picking is to be executed.

By defining a wall at some depth relative to the Leap Motion, the user can execute picking when perforating the defined wall. This does not require a quick gesture from the hand. By mediating the distance to the wall through the size of the elements depicting the finger screen positions, as well as changing the colors while perforating the wall, the user can slowly pick with a greater precision. This allowed the implementation of the mouse like functionality, with click; the frame the user perforating the wall, down; every frame the finger is behind the wall, and release; the frame the finger leaves the wall.

Figure 3.9: This illustrate the picking controls. By moving the hand through the metaphorical wall, picking is activated. When moving the hand away from the wall,

picking is released.

(40)

The interface areas affected by the new picking system were selection, menu interaction and gizmo activation.

Selection Selection of primitives is now executed through the release functionality of the picking, i.e. when leaving the defined wall after perforating it. The screen position of the release acts as the selection point. Rectangle screen selection is executed by perforating the wall at two different positions. When either position is released, the screen selection is executed between the two different positions.

Menu interaction Picking in the menu involves activating the different buttons of hte menu. Here the click functionality was mapped to the activation routine.

Gizmo activation Activation of the gizmos was mapped to the click functionality of the fingers. When perforating the wall while hovering the gizmo handles, the activation was executed. While activated, the gizmo followed the screen position of the hand until the release of the gizmo. The release functionality was still mapped to the key tapping gesture. This enabled the user to focus on the active transformation instead of the position of the hand. If the release of the gizmo was mapped to the release of the finger activating it, the user could have problems with unintended releases while transforming the gizmo.

Menu Layout

One thing that could be optimized was the menu interaction. While the menu interaction was one of the areas where the participants had the least problems, finding the correct buttons was observed to be a bit too time consuming. The menu layout was changed to let sub menus gather around the position of the parent button. This way the users did not have to traverse the whole screen to select buttons in sub menus, reducing the time spent in the menu.

Finger Occlusion

The redesign of the picking functionality resolved some of the problems that occurred because of the occlusion problem. With the implementation of visual feedback when picking, the user was able to see if the finger executing the picking disappeared from the screen unintentionally (was occluded). The problem with the camera transition was resolved through the use of thresh holds. Five or more fingers has to be visible to enter the camera transformation state. But to change state upon entering camera transition, a maximum of two fingers has to be visible.

This generates robustness to the controls.

(41)

Chapter 4 Method

In Chapter 3 a gestural interface was defined based on previous research and a pilot study. This chapter uses that interface to gather the data needed to answer the thesis research questions. The devloped interface was used in an experiment to evaluate the impact of implementing a gestural interface in a modeling tool, by letting human participants test the interface against the standard interface of the target modeling software. Information about modeling speed and the resulting quality of different tasks was gathered in the experiment and the implemented gestural interface was also evaluated. The latter is done through a System Us- ability Scale, SUS, test [21]. The user case study was designed to evaluate all the aspects of the virtual environment, the modeling tool, as stated by Bowman [7]; navigation, selection and manipulation. The implemented functionality is also focused on in the experiment. A first experimental design was created, and the content and execution of the experiment was evaluated with a pilot study.

Changes based on the results of the pilot study was implemented. Lastly, the execution flow of the experiment is presented.

4.1 Why Experiment

The main idea for experimentation is to match ideas with reality. The idea behind this thesis is to evaluate if the intuitiveness given by the gestural interface will ease the introduction of new users to modeling tools. By implementing a gestural interface for modeling based on previous research and then evaluating this against simulation of real situations, experimentation with human participants will help evaluate the interface; both through the time and quality, as well as the percieved usability of the interface.

Through the use of experimentation, the human factor is integrated into the results. With the combined interest of quantitative data, like time and resulting quality, and qualitative data through the perceived experience, the experiment will be able to gather both types of data.

32

(42)

4.2 Experiment Design

The experiment was designed as a two-step process. As with the developed interface, the experiment design was first developed to be evaluated in the pilot study. This to ensure that the content of the experiment is plausible to execute within the time of the experiment and that the data is obtainable through the experiment. The initial design was based on the data that was to be gathered.

4.2.1 Data

Here the data that was gathered to be able to answer the research questions of this paper regarding effectiveness in relation to speed and quality are listed. A definition of the data will be given in Chapter 5. The effectiveness is related to the time needed to execute the content of the experiment. The experiment was defined by a set of tasks, where each task was executed once for each interface.

Each task execution was timed and recoded from the start to the finish and acted as the final result.

The quality aspect is the produced results of each of the experiment tasks. As the platform for the interfaces was Autodesk Maya, the experiment was executed in scenes of the Maya software. These produced scenes contained the quality data that was retrieved from the experiment.

Finally, the two interfaces were evaluated based on the SUS test. This test was used to argue if the implemented interface is a good representative for the produced data as well as to be able to quantify the experienced usability and learnability of both of the interfaces.

4.2.2 SUS - System Usability Scale

SUS is a questionnaire that measure usability of softwares or websites. Created by John Brooke in 1986 it quickly became an established measurement for usability.

The SUS questionnaire consists of ten questions. See Appendix B.5. These questions are graded 1 - 7, where 1 refers to not agree at all, and 7 agrees totally.

Usability is defined as the intersection between the satisfaction, efficiency and effectiveness. SUS claim to measure the satisfaction aspect of the system [21].

The reason for doing this measurement was to determine if the implemented interfaces were satisfactory for the participants, and how they measured against each other.

4.2.3 Participants

The target participants of the experiment are to be unfamiliar with modeling tools. While there is a need to understand the concept of 3D and virtual environments, the participants should be new to the process of modeling. This way the

Interaction with a 3D modeling tool through a gestural interface: An Evaluation of Effectiveness and Quality