Materials

For this study a 2800 Samsung U28D590 ultra HD LED monitor was used to display the LDR images at resolution 3840_⇥2160 while a Dell UltraSharp 2007WFP 1900 LCD monitor was used to display the GUI. All the sounds were delivered binaurally using a set of Sennheizer HD 380 PRO headphones.

The distance from the participant’s head to the main display and the rotation angle needed to watch the contents of the secondary display are shown pictorially in Figure 7.2. These follow the guidelines of the ITU-R BT.500-13 standard [ITU12] for adjusting viewing conditions in subjective evaluations using HDTV and SDTV display panels. The experimentation procedure was conducted under dark ambient luminance conditions in a silent room.

A total of six scenarios were used for the experimentE1. These are labeled: Bathroom, Car, Kitchen, Kitti, Restaurant and Yard. The Bathroom scene was used only for training before the formal experimental session commenced and it was not used for collecting experimental data during this experiment. Images of the six scenarios are given in Figure 7.1.

The sound stimulus in the Kitchen scenario was assumed to originate from a kettle that was visible to the subjects. The Yard scenario consisted of the sound of

Figure 7.1: All the visual images of the scenarios used in E1 experimental study. FromLefttoRightand ToptoBottom: Bathroom, Car, Kitchen, Kitti, Restau- rant, Yard.

the engine ignition of a visible lawnmower at the right hand side of the screen. The Kitti sound stimulus was a hymn where the sound source originated from behind the main church tableux. For the Restaurant scenario the sound of an employee’s voice delivering food behind the wall was used while there was clatter noise coming from the same position. The Car scenario sound includes both engine ignition and the putting on of a seat belt. The Bathroom scenario included the sound of water flowing in the shower. The objective was to present both indoors and outdoors scenes where the audio stimulus varied from mechanical everyday sound to human speech and melody.

A custom GUI application was developed for the subjective studyE1. The interface included two main windows, one displayed the images in full screen at the main monitor and the other contained a window with the basic controls for adjusting

Figure 7.2: Top: Hardware set-up of theE1 experimental study. The viewing dis- tance to the main monitor and the angle needed to see the contents of the secondary monitor are given in [ITU12]. Bottom: Photo taken during preliminary tests.

the graphics and acoustics quality. The interface design was based on a pilot study with six participants. Their feedback assisted to improve the GUI before starting the experimental study. The participants of the pilot study did not take part in the experimental study for avoiding bias on the results. Two slider thumbs were used to dynamically change the quality of the presented visual and auditory stimuli by taking values that correspond to di↵erent quality levels.

The e↵ect of di↵erent budget sizes is shown in the GUI by colouring red the portions of the slider bar which contain quality levels that cannot be presented with the given budget. An example of the GUI configuration is given in Figure 7.3. This image presents an instance of the medium budgetB3and an instance of the smallest

budget size B1 used while it depicts all the controls that composed the GUI. The

visual slider bar, along with its label, were randomly interchanged with the auditory at di↵erent experimental trials to avoid participants’ adaptation to control only one of the bars for the task of allocating resources.

At the beginning of each trial the two sliders were located at a “null” stimulus mode, namely, at the left most point of the two slider bars. This configuration includes the presentation of a zero cost image and a zero cost audio to the subjects. A grey image was utilised as a zero cost image while a silent track was the zero cost audio track. Grey images are frequently used in experimental studies as a mean of neutralising participant’s eyes before the next experimental trial [ITU12]. The two thumbs always reset to the beginning of the bars when the user goes to the next trial. This is not the case when the trial includes the distribution of budget B5.

In that case, both thumbs start from an audio-visual level that has theoretical cost equal to B5 B4 and corresponds to a medium quality audio and image. Using

this thumb configuration the user was able to start exploring quality levels from the beginning before deciding the desired quality.

The movement of the two thumbs is independent at the beginning until the sum of the costs at the selected audio-visual levels exceeds the budget given for the trial for the first time. In that case, the last moved thumb is placed at the correct location so as the costs sum up to the total budget and the two thumbs become dependent on one another. The dependency of the sliders remains until the user decides to move on to the next trial. Participants become aware of the thumb dependency when they notice the synchronous movement of both slider thumbs.

The transition from independent to dependent GUI controls (thumbs) was decided on purpose so as the user can realise the constraint of the finite budget amount in VE computations. If the thumbs were always independent of one another, the task of resource allocation would be pointless as the users would have the option

Figure 7.3: Snapshots of the experimental software used including the GUI controls (buttons, slider thumbs, progress bar, etc.). Top: An instance of the budget size

B3. Bottom: An instance of the smallest budget size B1.

to maximise both audio and visual qualities at the same time. In that case the sum of the audio-visual costs would never be constant and the budget size could take values other that the ones given in table 7.1. This means, that the budget would not be an independent variable. Furthermore, the “next” control button remains disabled as long as the two sliders are still independent and is automatically enabled when they become dependent for the first time. This way, the user had to allocate all the available budget before going to the next trial.