2.1.

Objective

The user test in Experiment I is aimed investigating the ideal luminance contrast between the text and background for reading on smartphone display, as well as developing a model of the adaptive luminance contrast which enhances both reading performance and visual comfort.

2.2.

Stimuli

Prior to the experiment, stimuli for the user test were prepared. In creating luminance contrast, three approaches were employed: first, set A was text change—the text color changes from black to white as time passes while the background color is fixed as white; second, set B was background change—the text color remains black and the background shifts from white to black; and lastly, set C was text-background change—the text gradually changes from black to white, whereas the background color changes from white to black at the same rate. Next, given that the maximum luminance contrast between the text and background is 100 and the minimum luminance contrast is 0, each set was equally divided into 10 levels of luminance contrast. After combining the three overlapped stimuli from each set, black text on white background into one stimulus (hereinafter cited as default), there were a total of 28 luminance contrasts. The luminance was measured using a spectroradiometer (Konica Minolta CS-2000) under the same luminous environment during the experiment. Moreover, the display RGB input and Michelson contrast¹⁰ of each stimulus were listed in Fig. 1.

Fig. 1

RGB input, luminance and Michelson contrast of the 28 stimuli for Experiment I.

Among the 28 stimuli, some could cause negative perception in terms of both cognitive comfort and esthetics due to a very small luminance contrast between the text and background. Hence, before the user test, subjects were asked to indicate the stimulus with the smallest but still acceptable luminance contrast for each set. Based on the responses, the minimum luminance contrasts for sets A, B, and C were found to be T40, B70, and TB40, respectively (see Fig. 1). In addition, an intermediate stimulus was inserted for each set—T70, B80, and TB70—and seven stimuli were selected for the reading task.

2.3.

Experimental Setup

A total of 50 subjects composed of 26 males and 24 females took part in the user test. The average age of the subjects was 23.18 years with a standard deviation of 1.99 years. Each of them had normal vision or corrected to normal vision. The stimuli were displayed on a smartphone (Samsung Galaxy S3) with a 4.8 inch screen, and the backlight luminance of the display was fixed at $140\mathrm{cd}/{\mathrm{m}}^2$. At the subject’s desk, the measure of the correlated color temperature was 6300 K, and the illuminance was 600 lx. During the test, the subjects maintained a typical viewing distance from the smartphone display of 30 cm.¹⁸ The experimental setup is schematically shown in Fig. 2.

Fig. 2

Experimental setup for the user test: (a) displayed stimulus, and (b) experimental session.

2.4.

Method

Subjects were instructed to read 10 pages (approximately 2500 letters) of a newspaper article at a natural speed on a smartphone under the seven selected stimuli in a random order. They had a 1-min break before turning to the next reading sessions. Each article consisted of different content but the same level of difficulty to circumvent the influence of content type. In the same way as related studies for examining the effect of displays,^19–22 this study evaluated both the reading performance and visual comfort by conducting two kinds of tasks. First, to evaluate reading performance, the subjects’ reading speeds for each page were measured under the seven selected stimuli on the assumption that reading speed is faster under an adequate reading environment. Next, to assess visual comfort, subjects answered a self-report questionnaire to evaluate their subjective visual comfort to luminance contrast for the 28 stimuli by using a five-point Likert scale²³ (1 point: very uncomfortable; 5 points: very comfortable). For example, they gave a score of five points if they felt comfortable with the given stimulus.

2.5.

Results and Analysis

There was a large difference between the reading speeds of each page. It became faster as the subject turned the pages. The reading speed noticeably increased from page 7 or about 135 s after the subject began reading (average reading speed of the 1 to 6 pages: $672.07\text{letters}/\min$; 7 to 10 pages: $743.08\text{letters}/\min$), and the increased speed was thereafter maintained. This means that subjects generally start focusing attention on reading around 135 s after they begin. From this, the starting time of the change in luminance contrast could be determined given that reading speed does not change after the subject concentrates more intensively on the reading.

Comparing the reading speed of each set, two stimuli in set B (B80 and B70) recorded the highest speed and stimuli in set C (TB70 and TB40) followed closely, whereas a markedly lower speed was observed for the default stimulus (D100), as shown in Fig. 3. In other words, the stimuli in two high-ranked sets have a better reading performance than the others. The results were contrary to earlier research results stating that reading speed is fastest when the luminance contrast between text and background is maximal.^24,25

Fig. 3

Mean reading speed and visual comfort for the seven stimuli in Experiment I.

In terms of visual comfort, subjects felt most comfortable under the default stimulus (D100). Comparing visual comfort among two stimuli in same set and the default (D100, T70, and T40; D100, B80, and B70; D100, TB70, and TB40), visual comfort decreased rapidly in every set when the luminance contrast became smaller. In comparison with each set under same level of luminance contrast (intermediate and minimum luminance contrast), respectively, the stimuli in set C (TB70 and TB40)—the luminance of text and background change simultaneously—earned the highest comfort score in both times. Furthermore, in set C, the degree of visual comfort decrease that depended on the decline of luminance contrast was relatively small in sets A and B. On the contrary, B80 included in set B received a very low comfort score even at the intermediate level of luminance contrast. From the results, it is found that subjects do not feel visual comfort under the stimuli in set B despite the fact that those offered the best reading performance; as such, it is inappropriate to select set B as the changing pattern of luminance contrast.

2.6.

Development of Adaptive Luminance Contrast

By using the results of the user test, an empirical basis for designing the adaptive luminance contrast was partially collected. It is appropriate that the luminance contrast starts from maximum contrast (corresponding with the default) and gradually decreases in the text and background at the same rate, as in set C.

Accordingly, the recognition point of change in luminance contrast was investigated in order to confirm the final luminance contrast and to derive the adaptive model. A video clip with changing luminance contrast from maximum (0.70 in the experiment) to 0 for 20 s was created for set C, and subjects were asked to stop the video clip when they noticed a change in the luminance contrast on the display. As a result, it was found that the average recognition point was a luminance contrast of 0.52 in the experiment (luminance of text: $62.06\mathrm{cd}/{\mathrm{m}}^2$; luminance of background: $198.64\mathrm{cd}/{\mathrm{m}}^2$), namely the RGB values of the text are 51 and those of the background are 204; this remained the constant even though the changing time varied. Luminance decrease should occur within the imperceptible range because esthetic satisfaction of the display decreases if users notice the change.

Through a series of user tests, a time-dependent adaptive model of luminance contrast for a smartphone display was developed as shown in Fig. 4. The luminance contrast between the text and background starts from a maximum where RGB values of the text are 0 and those of background are 255, because a large contrast helps users to concentrate on reading at first. They also feel visual comfort for these values. However, users suffer from visual fatigue if it lasts a long time; hence, the luminance contrast begins to change 150 s later as they generally focused more intensively on reading the content at about that time. In addition, by taking into account both reading performance and visual comfort, set C where the luminance of the text and background change at the same rate was selected as a changing pattern. The final luminance contrast was determined as 0.52, the RGB values of the text are 51 and the background are 204 considering the average recognition point of change. In order to avoid users becoming aware of the change because they are bothered with the changing luminance, the luminance contrast between the text and background shifts slowly for 40 s.

Fig. 4

Adaptive luminance contrast: gradual decrease between text and background as time passes.

3.1.

Objective

The purpose of the second experiment of the study is to apply the adaptive luminance contrast in an actual usage environment. Moreover, it attempts to verify the effectiveness and applicability of the proposed adaptive model through a series of multidimensional validation tests by evaluating reading performance and visual comfort.

3.2.

Stimuli

Besides the adaptive luminance contrast, two additional luminance stimuli were prepared to compare the effectiveness with the developed model. The first stimulus was black text on a white background which corresponds with D100 in Experiment I, the luminance contrast of the current displays on general smartphones. Another stimulus was B70, which was assessed as the best reading performance from the previous user test. Thus, a total of three stimuli were comprised for the validation test as shown in Fig. 5.

Fig. 5

Mean ratio of high beta rhythm, reading speed, and visual comfort for the three stimuli in Experiment II.

3.3.

Experimental Setup and Method

Twelve students, consisting of six males and six females, served as subjects for the experiment. Their average age was 24.83 years with a standard deviation of 1.46 years. The experimental environment was identical with that of the user test in Experiment I.

Subjects were asked to read a newspaper article for 5 min using a smartphone under the three luminance stimuli in a random order. As in the previous test, both reading performance and visual comfort were evaluated to compare the effectiveness of the stimuli. In addition to measuring the subjects’ reading speeds and their self-reports on visual comfort, a brainwave analysis was carried out using an electroencephalogram (EEG) while subjects read articles. A total of four electrodes, two channels on the frontal lobe (${\mathrm{F}}_{\mathrm{P}1}$, ${\mathrm{F}}_{\mathrm{P}2}$), ground, and reference electrodes, were attached to the subject’s head to capture the EEG signal, as shown in Fig. 6. Brainwave data were measured for 5 min and 10 s, and the first 10 s of the recorded data were deleted to improve the accuracy of the results. The collected data were converted by 256-Hz sampling frequency and 12-bit analog-digital converter, then saved through Telescan, the analysis program developed by Laxtha Inc. After that, the relative ratio of the high beta rhythm (20–30 Hz)²⁶ to the entire range (3.5–50 Hz) was then calculated using the data. A high beta rhythm is generated when people feel stress and anxiety;²⁷ thus, a low ratio of high beta rhythm indicates that the subject is in a comfortable state during reading. Each subject’s data was obtained by calculating the mean ratio of two channels. For example, if the ratios of high beta rhythm from ${\mathrm{F}}_{\mathrm{P}1}$ and ${\mathrm{F}}_{\mathrm{P}2}$ were 7.11% and 6.98%, respectively, the mean value of the two data, namely 7.05%, was the final data for the subject. The relative ratio is used for brainwave analysis rather than absolute power since brainwave powers have wide variations in different people.

Fig. 6

Experimental setup for validation test: (a) attachment location of electrode and (b) experimental session.

3.4.

Results and Analysis

Through the results of the validation test, the effectiveness of the three stimuli was compared in various aspects, as shown in Fig. 5. The ratio of the high beta rhythm, which indicates the degree of stress during reading, was much lower when viewing the adaptive luminance contrast than when reading on D100, which corresponds with the luminance contrast of current smartphone displays. While reading under the developed adaptive model, the emission of high beta rhythm was about 7% and was even lower than that while reading books on paper²⁸ or working cognitive performance tests.^29,30 Moreover, the result of the self-report questionnaire on visual comfort also rated highest in the adaptive luminance contrast among the three luminance stimuli. That is, users feel most comfortable with the display applying the proposed model. On the contrary, B70 was reported as the least comfortable stimulus, because the dark background gave subjects an unpleasant impression, consistent with the user test in Experiment I.

In respect to reading performance, subjects read the article most quickly under the adaptive luminance contrast, followed by D100 and B70. This demonstrates that the developed adaptive model improves the readability of typography.

To sum up, the adaptive luminance contrast was the most outstanding for both reading performance and visual comfort, and it confirmed that a gradual decrease in luminance contrast provides an adequate reading environment on smartphone displays.