1. Introduction

Increasingly, spectrally tunable LED products and interactive lighting control systems allow users to change the colour of lighting without energy-wasting filters. Benefitting from advances in technologies, new uses for coloured light in architectural spaces and interior environments are being explored [1]. The potential advantages of varying the spectral power distribution (SPD) of emitted light goes beyond the decorative use of colours. The spectrum of light could be customised for museums artefacts to minimise photochemical damage or to enhance the faded colours of works of art [2, 3]. The colour rendering of skin tones can also be improved by the use of LEDs in the theatre [4].

After the discovery of intrinsically photosensitive retinal ganglion cells (ipRGCs), numerous studies have demonstrated that light significantly impacts circadian rhythms as a function of wavelength [5]. Spectrally tunable LEDs and lighting control products can allow the colour of light to be changed to enhance human health. Tunable LED products, which can theoretically provide any colour of light, are replacing lighting products that simply provide static white light [6].

Some lighting colour control systems have been commercialised, with varying levels of success. Although an understanding of usability should underlie the design of tunable LED systems, little research has been done on this. Some large manufacturers may conduct studies when developing and launching products, but results are presumably confidential and not disseminated. Anecdotal reports suggest that artistic colour spaces might be more easily understood by non-experts than scientific chromaticity systems, but this has not been studied empirically.

In colour science, scientists and engineers use colorimetry to quantify and predict chromaticity in specified colour spaces, such as CIE 1976 L*a*b* for object colour and CIE 1931 (x, y) for light colour. However, it appears as though manufacturers design control interfaces that are directly related to the relative intensities of multiple single-colour LED channels. Spectral tailoring applications may require light chromaticity to be precisely controlled. A conversion model could bridge the gap between research findings in colour science and the commercial products. In this research, a model is developed for transferring lighting device-specific control signals to chromaticity information in scientific colour spaces, which is then used to investigate the usability of various colour control interfaces. The usability of a control interface is impacted by a number of variables, such the graphic design, information visualisation and hardware elements (e.g. buttons, knobs, or touch panels). However, the objective of this experiment was to understand the impact of the colour mixing system only, not the visualisation design or the device development. It was designed to compare three colour spaces underlying any interface graphic design and hardware design.

2. Previous studies

For colour television and computer displays, calculation procedures have been developed for transforming chromaticity information among different colour spaces and for converting image information to the signals that control the display primaries [7]. Schwarz et al. noticed that, although displays provide 16 million colour options, users had difficulty specifying their desired colour. They conducted an experiment to compare the usability of RGB (red, green, blue), YIQ (luma, in-phase, quadrature), LAB (CIE 1976 L*a*b*), HSV (hue, saturation, value) and opponent colour models for displays [8]. The results suggest that inexperienced users complete colour matching tasks more rapidly with the RGB control interface, but less accurately, than with other colour models [8]. However, recent research shows conflicting results. Beigpour and Pedersen compared user performance for matching the chromaticity of a Philips Hue lamp with a chromaticity displayed on a computer screen, using RGB and HSV interfaces. In this experiment, participants matched colours faster, but less accurately, when using the HSV colour control interface [9].

The results from these experiments cannot be directly applied to lighting control systems. In both experiments, colours were presented on computer displays, which are self-luminous objects, for colorimetry purposes. In colorimetry, there are some important differences between light source and object colour. In particular, objects can appear brown or grey, but true light sources cannot. Additionally, objects have a dimension of lightness, whereas light sources have a brightness dimension. As a result, the colorimetric formulae used for self-luminous objects are different from those for the chromaticity of light [10]. Thus, this research investigates the usability of lighting control systems based on different colour control interfaces: RGB (red, green and blue), HSB (hue saturation and brightness) and opponent colour.

3. Methods

The experiment was conducted in the Lighting Laboratory at the University of Sydney, which is a windowless space painted black from floor to ceiling to control the light during the experiment. As shown in Fig. 1, two booths, one reference and one test booth, were constructed of identical white panels and illuminated with identical lights – ETC Source Four LED Profile x7 Color System. The dimensions and viewing distance are shown in Fig. 2. A neutral-density filter (ROSCO E Colour 211 0.9ND) was mounted in front of each light to reduce the intensity of light of all wavelengths. Each light has seven different LED channels, but only red (peak wavelength: 633 nm), green (peak wavelength: 523 nm) and blue (peak wavelength: 445 nm) channels were used in this experiment. The channels were controlled by the DMX values spanning 0-255. For instance, when the DMX value of the red channel was 255, the intensity of the red light was at its maximum.

 

Fig. 1 Experimental setup. The left booth was the reference booth and the test booth was on the right. Participants controlled the light in the test booth to match the light in the reference booth. General laboratory lights were only on to capture the photograph.

Download Full Size | PDF

 

Fig. 2 Plan view of the experimental setup.

Download Full Size | PDF

3.1 Light measurements

The SPD and luminance of a near-perfect reflecting white object, placed in the centre of the spot of the light in each booth, were measured when illuminated by each of the three channels, in steps of 25 DMX units, with a Photo Research Spectrascan PR-730 spectroradiometer. The measurements were done in both the test and the reference booth. Results showed that the SPD differences between the reference and test lights were small enough to be ignored, while the test and reference luminances were slightly different. Figure 3 shows the SPD measurements of the test light and Fig. 4 shows the luminance as a function of the DMX values for each primary light in reference and test booths.

 

Fig. 3 Relative power as a function of wavelength for the light used in the test booth. SPDs measured at every step (25 units) of the DMX values for each channel.

Download Full Size | PDF

 

Fig. 4 Luminance as a function of the DMX values for each primary light. Solid lines with triangle markers indicate the light used in test booth, while the dashed lines with circle markers indicate the light used in reference booth.

Download Full Size | PDF

3.2 Colorimetry computations from control signals and interface development

In order to develop functional colour control interfaces to use in the experiment, the relationship between the colorimetric and photometric parameters of any coloured light and the device-specific control signal information (DMX values) needed to be understood. Figure 4 shows that the luminance of each primary increases proportionally as the DMX value increases. Coefficients were computed for each LED channel to convert any DMX value to the luminance of that channel. Since the LED luminaires used in two booths were slightly different, the conversion formulae were set separately for each booth in the program, according to the coefficients obtained.

CIE colour matching functions (CMF) [11] and the SPDs obtained from the measurements were used to calculate tristimulus values, as shown as Eq. (1) [10].

Where ${{\displaystyle \varphi }}_{\lambda }(\lambda )$ is the SPD and $\overline{x}(\lambda )$,$\overline{y}(\lambda )$and$\overline{z}(\lambda )$are the CMFs.

The integration was carried out by numerical summation at a wavelength interval of 1 nm. According to Grassman’s Laws, additive colour-mixing is mathematically simple [12]. If A_r, A_g and A_b refer to the amount of the three primary lights, then tristimulus values of the mixture light (X_m, Y_m, Z_m) can be obtained by Eq. (2).

where X_R, Y_R and Z_R are the tristimulus values of red light, X_G, Y_G, Z_G are the tristimulus values of green light, and X_B, Y_B and Z_B are the tristimulus values of blue light.

Using the data from Fig. 3 for each primary light, the tristimulus values obtained from Eq. (1) can be expressed as a function of DMX values for each LED channel. The proportion of each of the three primaries can also be expressed as a function of the DMX value. Thus, the tristimulus values of any light mixed from this three-channel tunable LED luminaire can be predicted from the DMX values, as expressed in Eq. (3).

The 3 by 3 matrix (A) of constants was obtained by the computation using the data from Fig. 3. Then, chromaticity coordinates x, y in CIE 1931 (x, y) and u’,v’ in CIE 1976 (u’, v’) were calculated, using Eq. (4) and Eq. (5), respectively. Thus, DMX values were converted to chromaticity coordinates.

In this study, three different colour mixing interfaces were developed and tested. They were RGB (red, green and blue) colour mixing, HSB (hue saturation and brightness) specification and opponent colour mixing. Button boxes, as illustrated in Fig. 5, were developed to represent each of these systems and serve as the interface. These interfaces were very simple. The purpose of the button box was solely to provide an input mechanism for the increase and decrease of the experimental variables. The crudeness of the button box design would not be acceptable in a commercial product, but it was used to exclude the impact of the other design variables, such as the information visualisation, graphic design, etc. A touch-screen interface was not used because it would require a self-luminous device, which could interfere with the task.

 

Fig. 5 The button box used in the experiment. The control interface shown here was used for RGB control.

Download Full Size | PDF

The button box indicating the RGB control interface was labelled as shown in Fig. 5. When participants pressed the buttons from left to right, the lighting system increased and decreased the DMX values of the red channel, green channel and blue channel, respectively. Thus, participants directly controlled the intensity of each primary light. The resolution of each channel was normalised to the maximum intensity of each primary.

With the HSB interface, participants controlled the hue, saturation and brightness of colours. The HSB interface was designed based on CIE 1976 (L*a*b*) color space, abbreviated as CIELAB hereafter [13]. The button box was labelled to indicate ↑ (increase hue angle), ↓ (decrease hue angle), ↑ (increase saturation), ↓ (decrease saturation), ↑ (increase brightness), ↓ (decrease brightness) and “ENTER,” respectively. A Matlab program was developed to make real-time mathematical conversions between the DMX values and hue angles, normalised chroma and normalised brightness (based on luminance). Although an equal-energy radiator was selected in this experiment for the white point, and used to calculate X_n, Y_n, and Z_n values_, other light sources could be used. A program calculated all combinations of the DMX values resulting in (u’,v’) coordinates of (1/3, 1/3). Then, the combination of DMX values with these chromaticity coordinates and with the maximum luminance was identified as the reference white point, which had the CIELAB coordinates of (L* = 100, a* = 0, b* = 0). In this colour space, hue is specified by an angular measure (hue angle) and chroma (used to approximate saturation) is defined as the distance from the origin, excluding differences in lightness, as expressed in Eq. (6):

When a* and b* are both positive, the hue angle lies between 0° and 90°. When a* is negative and b* is positive, the hue angle is between 90° and 180°. The ↑ (increase hue angle) button rotated the hue angle anticlockwise, while ↓ (decrease hue angle) button rotated it clockwise. Chroma was used to approximate saturation (the colourfulness or vividness of the light). On the button box, the ↑ (increase saturation) button increased the chroma and the ↓ (decrease saturation) decreased it.

The opponent colour interface was based on opponent-colour theory, which posits colour perception is controlled by the activity of opponent channels [14]. Since activation of one end in the pair causes the inhibition of the opposite one, to decrease the proportion of one colour, such as red, participants were required to increase the proportion of its opponent colour, green. The first four buttons were used to increase the intensity of two pairs of opponent colours: red & green and yellow & blue. These buttons controlled the coordinates of the mixed colour along the green-red axis and the yellow-blue axis in CIELAB. The fifth and sixth buttons controlled the brightness (based on luminance) of the light.

These conversions could be generalized to commercial lighting products, including those using other digital control protocol signals. After SPD measurements of each channel and the calculations shown in Eqs. (1-6), manufacturers could relate any coloured light generated by their products to chromaticity coordinates.

3.3 Tasks and procedures

Eighteen subjects, seven men and 11 women, took part in the experiment. They were all aged 40 years or less and had normal colour vision, according to the HRR Pseudoisochromatic test [15]. None of the observers had any special training in lighting and did not work in the lighting industry. During the experiment, participants were seated approximately 1 m in front of the booths. Their task was to use each control interface to match the colour and luminance of the test light spot to the reference light spot, shown in Fig. 1. All participants had been briefly introduced to the nature and use of each of the three interfaces and had several minutes to practice before the experiment began. The instructions sheet that the participants read prior to commencing the experiment is shown in the Appendix. Some technical terminology has been replaced by layman terms to help participants understand. For instance, the luminance was referred as brightness. Buttons were labelled on the button box, as described earlier, according to which interface was being used. All three control interfaces were used by each participant throughout this experiment, but the sequence was randomised.

Six colour-matching trials, in which a participant matched six reference colours, were conducted with each control interface. There was no adaptation time between the trials, because, in actual illuminated environments, the adaptation state of occupants is not carefully monitored and controlled. Because of the randomised trials in this experiment, any negative impacts of adaptation would be expected cancel out. Furthermore, the aim of this experiment is to investigate differences in the data across colour mixing system conditions, and any impact of adaptation would affect performance with each of the three systems equally.

Six reference colours were specified in CIE 1976 (u’, v’) colour space, as shown in Fig. 6. At the end of each trial, participants pressed the “ENTER” button to indicate that they had finished a match, at which time the lights switched to the next trial automatically. Completion times and DMX values were recorded for each colour matching trial. The final DMX values were used to calculate the chromaticity difference and luminance difference between the reference and test lights. After completing six matches with each interface, participants answered a three-item questionnaire.

 

Fig. 6 Six reference colours and colour gamut plotted on CIE 1976 (u’, v’) chromaticity diagram. (Background colours for illustration only). Solid dots show the chromaticities of the R, G, B primaries used in this experiment. Any chromaticity within the colour gamut (the white triangle) can be obtained by mixing three primaries.

Download Full Size | PDF

3.4 Evaluation of usability

As suggested by ISO 9241 Part 11 (Guidance on usability), usability can be assessed by effectiveness, efficiency and satisfaction [16]. In this experiment, the effectiveness of reaching a desired colour with the control system was assessed by the accuracy of both luminance matching and chromaticity matching, which are inversely related to matching errors. Luminance matching error was the absolute difference between the test and reference luminances normalised to the reference luminance. Chromaticity matching error was quantified by calculating the Euclidean distance between the reference and test colours in CIE 1976 (u’, v’) chromaticity diagram, as expressed in Eq. (7) [10]:

Luminance and chromaticity coordinates were converted from the DMX values. The efficacy of the colour control interfaces is inversely related to the completion time, which was recorded during each task. Both effectiveness and efficacy are rather objective, but do not guarantee users satisfaction [17]. Jakob and Jonathan stated that, in 25% of their study cases, systems that the participants were most satisfied with were not the systems in which their performance was the best [18]. Thus, subjective assessment using questionnaires was also taken into account when evaluating usability.

3.5 Questionnaires

After the introduction, but before experiment starting, participants were asked: “If you had a chance to choose ONLY one interface to complete the colour-matching trials, which one would you choose?” This pre-experiment question aimed to understand users’ preference for these interfaces before actually using them.

After completing all matches for one interface, participants completed a three-item Likert questionnaire:

Seven-point scales were used to describe the level of their agreement or disagreement with each of the three statements. A rating of seven was the most positive response, indicating “strongly agree” while a rating of one indicated “strongly disagree” [19].

4. Results

Prior to beginning the experiment, 12 out of 18 (66.7%) participants believed that the RGB interface would be the easiest. This result is not surprising. The RGB interface appears to be more straightforward for inexperienced users than the other two and is likely familiar from the use of computer graphics software. However, results from both the psychophysical experiment and after-experiment-questionnaires contradict users’ initial opinion.

4.1 Control effectiveness

Chromaticity matching errors (Δ u’v’) for each trial were calculated using Eq. (7) and recorded. The mean chromaticity error and standard error of the mean (SEM) for each interface are shown in Fig. 7. Chromaticity matching errors with the RGB interface are significantly greater than with the other two interfaces. The results indicate no significant differences between the HSB and opponent colour interfaces.

 

Fig. 7 Chromaticity matching error (Δ u’v’) using different colour control interfaces. Circles show the mean Δ u’v’ and error bars show standard error of the mean (SEM).

Download Full Size | PDF

Figure 8 shows the luminance matching error (ΔL) with the different interfaces. The RGB interface caused the greatest luminance matching errors, while HSB interface led to the smallest errors. The mean luminance matching error with the RGB interface was 70%. With the HSB interface, the luminance difference was more reasonable, with a mean of 26%. The mean luminance difference with the opponent colour interface was 38%.

 

Fig. 8 Luminance matching error (ΔL) for different colour control interfaces. Circles show the mean ΔL and error bars show standard error of the mean (SEM).

Download Full Size | PDF

4.2 Control efficacy

Completion speed of the matching tasks, which is the inverse of the completion time, indicates the efficacy of a control system. Comparisons of speed or time in this experiment may introduce different biases against different interfaces. For instance, the RGB interface requires more button presses to adjust the luminance. Similarly, the efficiency of the hue settings with the HSB interface is dependent on the direct of travel selected by the participants. These issues, caused by the inherent nature of the interfaces, would also happen in real life. Thus, the comparison of completion time and speed can still be used to approximately evaluate the efficacy of a system. The completion time for each trial was recorded and the mean and SEM are shown in Fig. 9. When using the RGB interface, participants spent considerably more time to reach their desired colour than using the other interfaces. This indicates that the RGB interface was the least efficient interface among the three for the users to make their colour settings. The results showed no significant difference between HSB and opponent colour interfaces.

 

Fig. 9 Completion time (in seconds) of the matching task using different interfaces. Circles show the mean time and error bars show standard error of the mean (SEM).

Download Full Size | PDF

4.3 Questionnaire results

The Wilcoxon-Pratt test was used to analyse the subjective ratings collected from the after- experiment questionnaires. The Wilcoxon signed rank sum test is a classic test typically for non-parametric data, such as Likert scale ratings [20]. However, it has been criticised for excluding the zero observations before ranking, which makes the test perform poorly when there are serval zero observations in the data [21,22]. Pratt provided a modified version [22], know as “Wilcoxon-Pratt test”. Rahe generated a new critical value table for Pratt’s method [23].

Three interfaces were compared in three pairs in a Wilcoxon-Pratt test. The critical value of T in each comparison was obtained from Rahe’s critical values Table [23]. When the observed value of the Wilcoxon test statistic (T_obs) is smaller than the critical value of T, statistical significance of a difference can be established. The smaller T_obs is, the less likely the variation of subjective ratings of two interfaces is to have occurred by chance. Table 1 summarises the Wilcoxon-Pratt test results for each item on the after-experiment questionnaire. It can be concluded that users strongly prefer HSB and opponent colour interfaces over the RGB interface for all three items evaluating the perceived accuracy, perceived speed and overall user satisfaction. There is no obvious preference between HSB and opponent colour interfaces.

Table 1. Paired comparison of the subjective rating of three interfaces

View Table

5. Discussion

The comparison of the pre-experiment and after-experiment questionnaire is interesting. Before actually using the interfaces, most people (66.7%) believed that the RGB interface would be the easiest to reach the desired colours. However, in the experiment, it was the most difficult one. The RGB interface is often considered to be suitable for inexperienced users. The HSB and opponent colour interfaces are typically designed for professional users, such as lighting designers, since additional experience is presumably needed to operate them. For instance, the HSB interface requires users to be familiar with the sequence of hues on the colour wheel in order to quickly determine whether to move in a clockwise or anticlockwise rotation. The opponent colour interface requires the users to accept the concept that to decrease the proportion of one colour, one needs to increase the proportion of its opponent colour. However, this experiment demonstrated that inexperienced users were able to make their desired colour settings considerably faster with HSB and opponent colour interfaces, after a brief introduction, than with the more familiar RGB system.

One main difficulty when using the RGB interface is the control of luminance. Use of the RGB interface led to increased chromaticity matching errors, but the chromaticity differences, compared with the other interfaces, was not as great as the difference in luminance matching errors with the other interfaces. This suggests that the main drawback of the RGB interface is not in the control of chromaticity, but the control of luminance. As shown in Fig. 8, the mean ΔL with the RGB interface was 70%, but the mean of ΔL with the HSB interface was only 26%. In an earlier study with a similar experimental setup, but using a 3050 K white light, the mean illuminance matching error was found to be 18.5%, with the 95% confidence interval spanning 17.8% - 19.1%. This result was based on the data collected from 1440 trials [24]. In another experiment, in which participants viewed two lights (3050 K) directly and matched the luminance, the luminance difference was approximately 15% - 22% (95% confidence intervals) [25]. The RGB interface caused considerably higher luminance matching errors, presumably because few non-professional users can correctly adjust the luminance by controlling the intensity of three primaries, even though the method has been briefly described in the introduction. When using an RGB system, luminance co-varies with chromaticity, which can be confusing for some users.

Compared with earlier results of the white light experiments, the use of the HSB interface for coloured light has little negative impact on the luminance matching error. This interface appears to benefit from a separation in the control of luminance and chromaticity. The opponent colour interface led to greater luminance matching errors than HSB. This may be because inexperienced users cannot correctly distinguish the control of saturation and brightness when adjusting the output of the opponent channels. Users might adjust the luminance to change the saturation, which should be changed by adjusting the proportion of colours.

It is also worth noting that, regardless of which interface was used, the chromaticity matching errors were considerable greater than the just noticeable difference (JND). When Δu’v’ = 0.011, the chromaticity matching error is eleven times the just noticeable difference. It is not surprising that the acceptable chromaticity difference was much greater than detectable difference, since participants were told to complete the matching as accurate and as quickly as possible. In real life, users would likely not spend too much time with a control system to adjust the colour as well. Furthermore, in real illumination applications, users would likely be matching the colour of the light to a colour in their imagination, instead of to a neighbouring light source. The acceptable colour difference in real-life scenarios, with the use of a coloured light control system, would be expected to much greater than the JND. Further research to quantify the limits of acceptability would help to guide the design of the resolution of such a system.

Although the HSB interface is shown to be efficient, some improvements could still be made when designing an interface. To prevent the impact of additional light, self-luminous control panels were excluded in this experiment. Thus, to reach a desired colour, participants had to press buttons many times if the target and starting hue angles were far apart. In real life, this could be easily improved by employing some technologies, such as a touch screen. The button box used in this experiment was a very simple input device. Buttons are not necessary for HSB control. Instead, a colour wheel or swatch panel could be used to allows users to select a hue. Slider bars, knobs, or other components on a touch screen could be used to adjust chroma and brightness. The information visualization, underlying technology, and holistic user experience would need to be considered before developing commercial products.

6. Conclusion

Although participants may not necessarily be most satisfied with the system in which their performance was the best [18], users’ subjective ratings in this experiment were fully consistent with their performance, which leaves no doubt that the RGB interface has the lowest usability. The use of the RGB interface is time-consuming and ineffective – users were unable to reach their desired luminance. There was no significant preference shown in the experiment between HSB and opponent colour interfaces, but the latter did lead to greater luminance matching errors. The use of the HSB interface leads to the smallest luminance and chromaticity matching errors, suggesting this interface would enable users to easily reach their desired colour and brightness. New technologies could be combined with this interface to improve users’ experiences. Further studies could be conducted to investigate the magnitudes of perceptible and acceptable changes in light colour when using an HSB interface.

7 Appendix

Pre-experiment introduction to the use of three interfaces:

You will be seated and viewing two identical booths illuminated by two lights. Each LED light has three single-colour channels. By mixing these three primary colours, a large number of colours can be created. In this experiment, your task is to use the control interface to match the colour and brightness of the light spot in the test booth with the one in the reference booth.

Throughout this experiment, you will use three different control interfaces. The ways in which these interfaces control light are:

• • RGB (red, green and blue) colour mixing
• • HSB (hue, saturation and brightness) specification
• • Opponent colour mixing

A label on the button box will indicate which interface is being used. Each section includes six colour-matching trials. For each trial, press the ENTER button after you have made your match. Then, the experiment will switch to next trial automatically.

After completing six matches with one interface, you will complete a three-item questionnaire.

Description of three control interfaces

When using the RGB (red, green and blue) interface, you will control intensity of each primary colour. As shown in Fig. 10, the first two buttons control the intensity of red light. The up arrow button increases the intensity, while the down arrow button decreases the intensity. The third and fourth buttons control the intensity of green light. Similarly, the fifth and sixth buttons control the intensity of blue light.

 

Fig. 10 Label on the button box for RGB control.

Download Full Size | PDF

Light colour mixing is additive colour mixing, unlike the colour mixing of paints. For example, to get yellow light, you need to mix red and green light together. While mixing red and blue light, you will get purple light. To increase the overall brightness, the intensity of all three colours should be increased.

The HSB interface controls the hue, saturation and brightness of colour. Hue is sometimes referred to as “colour”, such as red or green. It is specified by an angular measure - hue angle. For example, you could choose a colour point on one of the colour circles in the Fig. 11. Then, the hue angle is the angle between that point and another point.

 

Fig. 11 Colour settings for HSB interface. (Background colours for illustration only).

Download Full Size | PDF

If starting at red, the hue for yellow will be 90° away, green would be 150° away, and blue would be 270° away. If you moved the full 360°, you’d be back at red.

Saturation refers to how colourful or vivid the light is. 100% saturation means the colour is completely pure, such as a vivid red colour. When the saturation is 0%, the light appears white. When it is between these values, the light may appear pink.

So, to get a pastel green colour, you would need to adjust the hue angle (if starting from red) to the green and then decrease the saturation.

On the button box labeled as Fig. 12, the first two buttons control the hue angle of light. The up arrow button rotates the hue angle anticlockwise, while the down arrow button rotates the hue angle clockwise. The third and fourth buttons control the level of saturation of the light. The up arrow button makes the colour more saturated while the down arrow button desaturates the colour. The fifth and sixth buttons control the brightness of the light. The up arrow increases the brightness while the down arrow decreases the brightness.

 

Fig. 12 Label on the button box for HSB control.

Download Full Size | PDF

In the opponent colour interface, there are two pairs of opponent colours: red & green and yellow & blue. The opponent-colour theory suggests that colour perception is controlled by the activity of opponent channels. The red-green channel exists because red and green are opposite colours. That is why we never perceive “reddish green” colours. Similarly, we neither see “bluish yellow”. Activation of one end in the pair causes the inhibition of the opposite one. Thus, to decrease the proportion of red light, you could increase the proportion of its opponent colour, green.

As show in Fig. 13, the first button increases the proportion of red light, while the second button increases the proportion of green light. The third button increases the proportion of yellow light while the fourth button increases the proportion of blue light. The fifth and sixth buttons control the brightness of the light. The up arrow button increases the brightness and the down arrow button decreases the brightness.

 

Fig. 13 Label on the button box for opponent colour control.

Download Full Size | PDF

Acknowledgments

The authors thank the anonymous peer reviewers for their comments, whose insight helped improve the quality of this paper.

References and links

1. S. O. Marberry and L. Zagon, The Power of Color: Creating Healthy Interior Spaces (Wiley, 1995).

2. F. Viénot, G. Coron, and B. Lavédrine, “LEDs as a tool to enhance faded colours of museums artefacts,” J. Cult. Herit. 12(4), 431–440 (2011). [CrossRef]  

3. D. Abdalla, A. Duis, D. Durmus, and W. Davis, “Customization of light source spectrum to minimize light absorbed by artwork,” in Proceedings of CIE Lighting Quality and Energy Efficiency Conference 2016 (Commission Internationale de L'Eclairage, 2016), pp. 22–31.

4. M. Wood, “Many shades of white,” Stage Directions 23 (9), (2010).

5. T. A. LeGates, D. C. Fernandez, and S. Hattar, “Light as a central modulator of circadian rhythms, sleep and affect,” Nat. Rev. Neurosci. 15(7), 443–454 (2014). [CrossRef]   [PubMed]  

6. W. Julian, “The Sydney opera house concert hall celebrates two years of LEDs,” Lighting 36(4), 26–27 (2016).

7. A. R. Smith, “Colour gamut transform pairs,” Comput Graph (ACM). SIGGRAPH 12(3), 12–19 (1978). [CrossRef]  

8. M. W. Schwarz, W. B. Cowan, and J. C. Beatty, “An experimental comparison of RGB, YIQ, LAB, HSV, and opponent color models,” ACM Trans. Graph. 6(2), 123–158 (1987). [CrossRef]  

9. S. Beigpour and M. Pedersen, “Color Play: Gamification for Color Vision Study,” in Proceedings of Color and Image, Midterm Meeting of the International Colour Association (AIC, 2015).

10. J. Schanda, Colorimetry: Understanding the CIE System (John Wiley & Sons, Inc., 2007).

11. C. V. R. Labs, Colour and Vision Database: Colour Matching Functions. Available: http://www.cvrl.org/(2016)

12. D. B. Judd, Colorimetry (U.S. Govt Print Off., 1950).

13. K. Witt, “CIE Color Difference Metrics,” in Colorimetry: Understanding the CIE system, J. n. Schanda, Ed., (John Wiley & Sons, Inc., 2007).

14. R. L. De Valois and K. K. De Valois, “A multi-stage color model,” Vision Res. 33(8), 1053–1065 (1993). [CrossRef]   [PubMed]  

15. I.-s. C. C. The Subcommittee on Color Blindness Studies, HRR Pseudoisochromatic Plates, (2002)

16. ISO, Ergonomic requirements for office work with visual display terminals (VDTs) (ISO, 1998)

17. M. Hassenzahl, “The Interplay of Beauty, Goodness, and Usability in Interactive Products,” Hum. Comput. Interact. 19(4), 319–349 (2004). [CrossRef]  

18. J. Nielsen and J. Levy, “Measuring usability: preference vs. performance,” Commun. ACM 37(4), 66–75 (1994). [CrossRef]  

19. J. R. Lewis, “Psychometric evaluation of an after-scenario questionnaire for computer usability studies- the ASQ,” SIGCHI Bulletin 23(1), 78–81 (1990). [CrossRef]  

20. F. Wilcoxon, “Individual comparisons by ranking methods,” Biom. Bull. 1(6), 80–83 (1945). [CrossRef]  

21. W. J. Conover, “On methods of handling ties in the Wilcoxon signed-rank test,” J. Am. Stat. Assoc. 68(344), 985–988 (1973). [CrossRef]  

22. J. W. Pratt, “Remarks on Zeros and Ties in the Wilcoxon Signed Rank Procedures,” J. Am. Stat. Assoc. 54(287), 655–667 (1959). [CrossRef]  

23. A. J. Rahe, “Tables of critical values for the Pratt matched pair signed rank statistic,” J. Am. Stat. Assoc. 69(346), 368–373 (1974). [CrossRef]  

24. W. Hu and W. Davis, “Dimming curve based on the detectability and acceptability of illuminance differences,” Opt. Express 24(10), A885–A897 (2016). [CrossRef]   [PubMed]  

25. W. Hu and W. Davis, “Luminance resolution of lighting control systems: usability and energy conservation,” in Proceedings of CIE 2016 ”Lighting Quality and Energy Efficiency (CIE, 2016), pp. 253–263.