5.1 Scene Development

There are two phases to this research. Phase one consists of replicating Mikellides experiment in virtual reality. Phase two aims to develop the experiment further by testing more colour variations, consisting of red, green and blue at both full and half saturation levels, to compare results between colour saturation and hue.

The virtual reality scene was developed to be as simple as possible to not interfere with the participants reactions to colour. The scene is plain room with the dimensions of 3.5✕4.5m with the walls, ceiling and floor all painted with the same colour and a light in the middle. The reason the room is square rather than a sphere where one would only be exposed to just the colour without shadows in the corners that change the colour slightly, is to maintain immersion and the feeling that the scene could be real like the experiment that this study is replicating. By being square the room has difference in shades of the colour especially visible in the corners but since all the rooms have the same lighting qualities this variable is controlled. The colours for the rooms in phase two are all on an RGB scale with changes in saturation (the addition of white into the colour). The chosen colours are:

TABLE 1. Colours used in VR

 

R100
Red value	Green value	Blue value
255	0	0
Saturation (%)	Hue	Hex code #RRGGBB
100	0	#ff0000
G100
Red value	Green value	Blue value
0	255	0
Saturation (%)	Hue	Hex code #RRGGBB
100	120	#00ff00
B100
Red value	Green value	Blue value
0	0	255
Saturation (%)	Hue	Hex code #RRGGBB
100	240	#0000ff
R50
Red value	Green value	Blue value
255	128	128
Saturation (%)	Hue	Hex code #RRGGBB
50	0	#ff8080
G50
Red value	Green value	Blue value
128	255	128
Saturation (%)	Hue	Hex code #RRGGBB
50	120	#80ff80
B50
Red value	Green value	Blue value
128	128	255
Saturation (%)	Hue	Hex code #RRGGBB
50	240	#8080ff

 

The reason these are the chosen colours is because they are universal and easily transferable to other screens since they are composed of LEDs consisting of RGB. Green was added to the original colours in the experiments of red and blue to add more variety rather than the usual cool versus warm colours, adding another dimension. In future tests a large variety of colours could be tested but for this research the main colours of RGB (Red, Green, Blue) that construct all other colours in the spectrum will be the only ones tested. To further test Mikellides conclusion that colour saturation has a larger impact than colour hue, the colours tested are RGB with 100% saturation and the same colours but with a saturation value of 50%.

 

The program used to create the VR environment was Unreal Engine 4 version 4.17, this was due to the programs ability to easily produce high quality scenes as well as its adaptability to create VR compatible content. To create a greater resolution to the scene, the lightmap resolution was modified to have the ideal texel density. The images show the room at 4, 52 and 250 texel density, 52 was the final value used for the rooms as it was the most efficient and effective at producing a realistic feeling spaces.

TABLE 2. Lightmap Resolution and texel density

 

Texel Density: 4	Texel Density: 52		Texel Density: 250
Insufficient texel density (4)		Ideal texel density (52)

 

Each room scene is exactly the same with just the colour of the material changing. The light is placed in the middle of the room providing the same lighting for all the walls. The material for the room had to be modified to not be reflective by changing the specular value to 0 to prevent a distracting hotspot on the wall.

 

Figure 1. Left, showing room before and right side the specular value in material was set to 0 to reduce light hotspot.

There were some problems encountered when building the lighting in Unreal Engine 4, all the lighting that was inside of the rooms would go black. The rooms were created by having two BSP’s, a smaller subtractive one within a larger one to create the cavity inside to ensure that it was a sealed room. These BSPs were then converted to static meshes for easier duplication and material assignation. It was found that the lighting problem was a result of using BSP brush since there is a problem with the current version of ue4 with dealing with bsps and lights. To fix this the rooms had to be remodelled through 3dsMax or by using a static mesh cube transformed to be each wall, ceiling and roof.

5.2 Timing

Mikellides’ experiment consisted of testing participants in the room for 20 minutes for each colour variation. For the replication of this experiment participants were also be exposed to the variations for 20 minutes each to replicate the experiment as faithfully as possible. Although for phase two of the experiment on extending the research into the effects that colour saturation has versus colour hue since there are more colours to test. To get a larger sample size, the time spent in each coloured room will be one minute. Accounting for a total of six minutes inside of the coloured environments. This is because one minute is the shortest time it could be taken to to still achieve required exposure time to get valid results. Exposure to the scene was tested at 30 seconds per room but this proved to be insufficient time for an appropriate reaction to be measured. The shorter the experience, the more likely participants would be willing to go through it since it’s easier to fit 15 minutes into the work routine than 1 hour and 10 minutes, especially in the busy work environment.

5.2.1 Iteration I

The first prototype of the scene, the order of the coloured rooms was the high saturation rooms first followed by the low saturation rooms i.e. R100, G100, B100, R50, G50, B50. The participant spent 40 seconds in each room. The results showed a contrast between the high saturation (HS) colours and the low saturation(LS) colours but since the HS colours were at the beginning of the experience, The decrease in reaction especially in the EDA results could be due to the saturation or because the participant has grown used to the experience therefore grows more relaxed as time progresses. To decrease the reactions being according to time relaxation, the coloured rooms were mixed up for the second iteration.

5.2.2 Iteration II

In the second iteration the sequence was: Start (S), R100, white room intermission (WI), G50, WI, B100, WI, R50, WI, G100, WI, B50. The time in each room remained the same as iteration one but interlude rooms were added in between each coloured room for 7 seconds. The interlude room was coloured a light gray/white since it is a neutral colour. The interlude room was also added to the beginning of the sequence so that participants have a moment to become accustomed to the dimensions of the room and the VR experience. The white interlude space did not have the desired effect on the tested participant. Rather than relaxing their eyes, the colour white was a stark contrast that agitated the eyes. In this prototype it was noticed that the current exposure time of 40 seconds would not be enough to achieve a valid mean, therefore for the next iteration the time was increased to 1 minute.

5.2.3 Iteration III

In the third prototype, the intermission rooms were eliminated since they were not helpful to the participant but the starting room remained since it provided time for participants to become accustomed and also allowed for leeway in case anything was wrong with the equipment such as loss of signal from the Bitalino or if there was a problem with the headset. The sequence was: S, R100, G50, B100, R50, G100, B50. The time in each room was one minute. The problem with this iteration was that the participants tested did not see the real rooms colour but an overlap of the previous colour with the new, this is explained further in section 5.3.

5.2.4 Iteration IV

The fourth and final iteration was a collection of the best aspects of the previous iterations. Resulting in a sequence of: S, R100, black intermission (BI), G50, BI, B100, BI, R50, BI, G100, BI, B50, S.

 

 

 

Figure 2. Sequence of Events for Virtual Reality environment.

5.3 Eye Colour Perception

During participant test of model in iteration 3, where the duration inside of coloured rooms increased to 60 seconds with no interludes, there was a change in colour perception for participants. This was most evident in the change from R100 to G50, rather than seeing the light green multiple participants saw a turquoise/aqua blue. This is due to the extended exposure to the red room, the brain is tricked into changing its perception of the light green. This problem is being fixed by providing an interlude time in between each colour change to give the eyes time to ‘reset’.

 

The interlude was tested as a white/light grey painted room but this did not provide the needed rest for the eyes since the effect of colours crossing over still took place. Therefore a black screen will be used as the reset interlude. The time needed for this interlude was tested by exposing participant to four different variations of interlude times and recording the colours that were perceived after the interlude.

5.3.1 Interlude of 5 seconds

TABLE 3. Annotations of testing an interlude of 5 seconds.

 

Colours in order	Actual Room Colour	Colours perceived
R100		Normal
G50		Bright aqua colour, towards the end of the minute it starts to fade slightly into the required colour
B100		Normal
R50		Normal but slightly brighter
G100		Normal but there are circles where the headset moves, as if caused by a mounted spotlight
B50		Normal but slightly stronger

 

5.3.2 Interlude of 10 seconds

While testing the environments a slight defect was found that the lighting intensity changes according to where the user looks, similar to having a spot light mounted on the headset. This spotlight effect was most evident in the full saturation green. This defect was fixed so that environment colours seen by the user are not modified.

 

“Finding that my eyes grow tired and am thankful for the black space. Although I feel like closing my eyes during this period but scared of missing the (room) transmission”

To fix this issue a sound was added at the end of each black intermission so if participants want to rest eyes by closing them they can until they hear the beep. This will also help in the signal data by providing markers for when it changes colour. At the end of the simulation the last scene is like the start, a white/grey room, a few sounds will play when they reach this stage to show that they have finished the experiment.

TABLE 4. Annotations of testing an interlude of 10 seconds.

Colours in order	Actual Room Colour	Colours perceived
R100		Normal
G50		No aqua but lighter than before
B100		Normal
R50		Normal
G100		Normal
B50		Normal

 

5.3.3 Interlude of 15 seconds

This interlude time provides enough time for eyes to readjust to normal. The light green room that was troublesome does not appear bright aqua anymore. Although while in the interlude from R100 to G50 with eyes open there is a slight blue tint on the black, similar to when one looks at a light for a bit then closes eyes there is a leftover colour where you were looking at the light.

TABLE 5. Annotations of testing an interlude of 15 seconds.

Colours in order	Actual Room Colour	Colours perceived
R100		Normal
G50		Normal
B100		Normal
R50		Normal
G100		Normal
B50		Normal

 

 

5.3.4 Interlude of 30 seconds

The 30 second interlude provide more than an adequate amount of time for the eyes to adjust although this time is longer than is necessary.

TABLE 6. Annotations of testing an interlude of 30 seconds.

Colours in order	Actual Room Colour	Colours perceived
R100		Normal
G50		Normal
B100		Normal
R50		Normal
G100		Normal
B50		Normal

 

In conclusion the time of the interlude will be 15 seconds since it provides the needed time for resting and ‘resetting’ eyes without taking up too much time of the overall experience. A short beep sound has been added to the end of the interlude times so that participants are warned when there will be a colour change in the room. This sound only appears at the end of the interlude time not at the beginning of the interlude, due to not wanting to saturate participants with the noise. The functionality of the noise is in case participants close eyes during the rest period they are warned to open them again when they hear the beep.

5.4 Sensor Data Analysis

The Bitalino has a complementary program called OpenSignals (r)evolution, this program is used during the recording of the signals providing a real-time sensor data recording. This data was then exported as a .txt file to be processed in Jupyter notebooks with custom made Python code.

5.5 Experiment Procedure

Participants were seated in a chair and a brief introduction of the equipment and experiment requirements was given along with a written explanation. The Bitalino board that was mounted in an adjustable arm band was placed midway on left arm, adjusting the position depending on the length of the arm so that the cables that extended from the Bitalino could reach the participant’s left hand and left side of the upper chest. Guidance was provided for the application of the ECG sensors on the left side of the chest and for the EDA sensors. Open Signals (r)evolution was used to record sensor data and a brief test before experiment commencement was done to check that signal outputs are correct.

 

Figure 2. Diagram of sensor placement.