Tiresias :: Scientific & technological reports

An Assessment of the Benefit of Internal Illumination in Keypads

by Antoinette Fennell

Introduction and Methods

The aim of this project was to investigate the effectiveness of internal illumination of keypads. In this study, two keypads are compared, one with internal illumination, the other without. The keypads are compared under three different ambient lighting conditions and three different levels of internal illumination are tested.

The Experimental Keypads

The first keypad was a twelve-key keypad with white labels on black keys (see Fig. 1). Three function keys are also located on the right hand side: a red cancel key, a yellow clear key and a green enter key.

Figure 1. Photograph of the non-illuminated keypad with randomised letter labels on the "numeric" keys.

Photograph of non-illuminated keypad with letter labels

The second keypad had an identical layout, but with black labels on white keys and with internal illumination (see Fig. 2a,b,c). The twelve "numeric" keys are illuminated from within with white light, so that the black label is silhouetted within the surrounding lit area. The cancel, clear and enter keys are internally illuminated with red, yellow and green coloured light, respectively.

Figure 2a. Photograph of the internally-illuminated keypad under dark ambient lighting with randomised letter labels on the "numeric" keys.

Photograph of the internally illuminated keypad with letter labels under dark ambient lighting condition

Figure 2b. Photograph of the internally-illuminated keypad under dim ambient lighting with randomised letter labels on the "numeric" keys.

Photograph of the internally illuminated keypad with letter labels under dim ambient lighting condition

Figure 2c. Photograph of the internally-illuminated keypad under bright ambient lighting with randomised letter labels on the "numeric" keys.

Photograph of the internally illuminated keypad with letter labels under bright ambient lighting condition

The test subjects were given the opportunity to use both keypads during the course of the testing procedure.

The Experimental Keypad Labels

A user would expect to find numeric characters on a twelve-key keypad. Therefore, even if the numbers were randomly arranged, the fact that they were present at all could in itself cause the user to press incorrect keys. Therefore to avoid bias, the "numeric" keys were labelled with letters (a = 1, b = 2, c = 3…. i = 9, j = 0).

To eliminate bias resulting from user familiarity with alphabetic order, the order in which the "numeric" keys were labelled was randomised.

Two different random letter arrangements were used for the two keypads, to reduce bias as a result of the user learning the letter order during the course of the testing procedure. See Figs. 1 and 2a,b,c for photographs of the keypad labels.

Internal Illumination Level

Three different levels of internal illumination were investigated. The illumination settings were 3V ('low'), 4.5V ('medium') and 6V ('high'). Each test subject only experienced one of the three internal illumination levels. The internal illumination level was alternated between test subjects, so that each subject only experienced one of the three.

Ambient Lighting Level

Three ambient lighting levels were used in this study. For purposes of analysis these levels were categorised as 'dark', 'dim' and 'bright'. The study was carried out indoors, so natural light could not be used. Also, the variation in natural light during the course of the day, and from day to day, meant that the level of natural lighting would not have been controllable.

In order to keep the ambient lighting level constant, the external windows (i.e. those allowing natural daylight in) were blacked out. The 'bright' ambient lighting level was 555 Lux (or 51.5 FC) and was created by turning on all of the lights in the laboratory. The 'dim' ambient lighting level was 8.4 Lux (or 0.79 FC) and was created by turning all of the laboratory lights off, and allowing the artificial hallway lighting through the internal windows. The 'dark' ambient lighting level was 0 Lux (or 0 FC) and was created by blacking out all of the windows, external and internal.

The test subjects experienced all three ambient lighting conditions.

Experimental Procedure

On entering the room, the test subject was asked to sit down. With the keypads out of view, the aims of the study and the procedure were explained.

One of the two keypads was placed in front of the test subject and the ambient lighting was set to one of three settings. A list of forty randomly sorted letters (from a-j) and function key labels (clear, cancel, enter) were called out by the evaluator in groups of four (e.g. "g, f, cancel, a", "clear, b, c, enter" etc.) and the test subject was asked to type the letter and function key combinations. After each group of four, the evaluator paused until the test subject had finished typing.

This test was repeated six times in total, covering all combinations of the two keypads (illuminated and non-illuminated) with the three ambient lighting conditions (bright, dim and dark). The order in which the six test conditions were carried out was randomised for each test subject.

Finally the test subject was asked to provide his/her comments and opinions on the two keypads.

Test Subjects

Nine test subjects in total took part in this study. Of these, six were partially sighted (four men, two women) and three had good vision (one man, two women). The test subjects were recruited from visitors and staff members of RNIB Bristol.

The test included subjects with a wide variety of eye conditions. Visual acuity was not tested during the course of this study; information on eye condition was provided by the subjects themselves. The reported eye conditions were Nystagmus (n = 2), extreme myopia with retinal detachment (n = 1), severe colour blindness (n = 1), Macular Degeneration (n = 1), and tunnel vision (n = 1).

Statistical Analysis

The data were analysed using SPSS Version 14.0 for Windows. Where data did not have a normal distribution non-parametric tests were used.

Six test conditions are named and defined as follows:

Non-dark: Non-illuminated keypad in 'dark' ambient lighting
Non-dim: Non-illuminated keypad in 'dim' ambient lighting
Non-bright: Non-illuminated keypad in 'bright' ambient lighting
Ill-dark: Internally illuminated keypad in 'dark' ambient lighting
Ill-dim: Internally illuminated keypad in 'dim' ambient lighting
Ill-bright: Internally illuminated keypad in 'bright' ambient lighting

The results of the six test conditions for partially sighted test subjects were compared with those of the test subjects with good vision using Mann-Whitney test for two independent samples. For the remainder of the analyses, both sets of results were pooled.

The two keypads were compared using the Mann-Whitney test for two independent samples. The three ambient lighting conditions were compared using the Mann-Whitney test for two independent samples. The three internal illumination levels were compared using One-Way ANOVA.

Finally, the six test conditions were compared using Friedman test and Wilcoxon Signed Ranks Test.

Hypotheses

Five hypotheses were tested in this study.

Hypothesis 1: Is there a difference in data input success for partially sighted test subjects versus test subjects with good vision? The null hypothesis states that there is no significant difference in the data input success for partially sighted test subject versus test subjects with good vision.

Hypothesis 2: Does internal illumination on a keypad improve data input success? The null hypothesis states that there is no significant difference in data input success between the non-illuminated keypad (see Fig. 1) and the internally illuminated keypad (see Fig. 2).

Hypothesis 3: Does ambient lighting affect data input? The null hypothesis states that there is no significant difference in data input success under 'dark', 'dim' and 'bright' and ambient lighting conditions.

Hypothesis 4: Does internal illumination level have an effect on data input success? The null hypothesis states that there is no significant difference in the data input success for the three different internal illumination levels ('low', 'medium' and 'high').

Hypothesis 5: Is there a difference in data input success on an internally illuminated keypad compared to a non-illuminated keypad, under different ambient lighting conditions? The null hypothesis states that there is no significant difference in the data input success for the two different keypads under different ambient lighting conditions.

Results

Hypothesis 1: Is there a difference in data input success for partially sighted test subjects versus test subjects with good vision?

For five out of the six testing conditions, there was no significant difference in the data input success between the partially sighted test subjects and the test subjects with good vision. The data input success for the internally illuminated keypad in 'bright' ambient lighting condition was significantly lower for partially sighted test subjects than for test subjects with good vision (see Table 1).

Table 1. Results of the pairwise comparison of data input success for partially sighted test subjects versus test subjects with good vision, for the six testing conditions, using data input success averages for each test subject.

Test Condition	Proportion Correct		Pairwise Comparison (Mann-Whitney Z)	Significance level
Test Condition	Partially Sighted (n = 6)	Good Vision (n = 3)	Pairwise Comparison (Mann-Whitney Z)	Significance level
Non-dark	0.00	0.00	0	n.s.
Non-dim	0.96	0.94	-0.9	n.s.
Non-bright	0.96	0.99	-1.6	n.s.
Ill-dark	0.99	0.96	-1.8	n.s.
Ill-dim	0.99	0.96	-1.8	n.s.
Ill-bright	0.96	1.00	-2.1*	*p < 0.05

Hypothesis 2: Does internal illumination on a keypad improve data input success?

With the data for the three internal illumination levels and the data for the three ambient lighting levels pooled, the data input success for the internally illuminated keypad was significantly higher than for the non-illuminated keypad (Mann-Whitney test: Z = -19.8, n = 2160 data input attempts, p < 0.001).

The test was repeated using the test subject averages and the result was equally significant (Mann-Whitney test: Z = -3.6, n = 9, p < 0.001).

Hypothesis 3: Does ambient lighting affect data input?

With the data for the two different keypads and the data for the three internal illumination levels pooled, the data input success under the 'dark' ambient lighting level was significantly lower than for the 'dim' (Mann-Whitney test: Z = -20.4, n = 1440 data input attempts, p < 0.001) and the 'bright' (Mann-Whitney test: Z = -20.7, n = 1440 data input attempts, p < 0.001) lighting levels. This test was repeated for the test subject averages and the results were equally significant (results not shown).

There was no significant difference between the 'dim' and the 'bright' data input success (Mann-Whitney test: n.s.)

Hypothesis 4: Does internal illumination level have an effect on data input success?

With the data for the three ambient lighting levels pooled, there was an increase in data input success as the internal illumination level increased from the low (3V) to the medium (4.5V), to the high (6V) level. The variation in the results was significant (One-Way ANOVA: F = 4.2, df = 2, p < 0.05).

When the test was repeated using test subject averages, the results were only nearly significant (One-Way ANOVA: F = 3.0, df = 2, p = 0.067).

For the pooled data, the data input success for the low level was significantly lower than for the medium (Mann-Whitney test: Z = -2.0, n = 720 data input attempts, p < 0.05) and the high level (Mann-Whitney test: Z = -2.6, n = 720 data input attempts, p < 0.05). The increase in data input success from the medium to the high level was not significant (Mann-Whitney test: n.s.)

Hypothesis 5: Is there a difference in data input success on an internally illuminated keypad compared to a non-illuminated keypad, under different ambient lighting conditions?

Six different test conditions were tested: a Non-illuminated keypad in 'dark', 'dim' and 'bright' ambient lighting and an internally illuminated keypad in 'dark', 'dim' and 'bright' ambient lighting.

A summary of the descriptive statistics for the six different test conditions is provided in Table 2.

Table 2. Descriptive Statistics of the Input Data for the six different test conditions for n = 9 test subjects.

Test Condition	Total data inputs N	Sum of correct inputs	% correct	S.E.	Mean Rank
Non-dark	360	0	0	0	1.0
Non-dim	360	344	96	1	3.2
Non-bright	360	350	97	1	3.9
Ill-dark	360	352	98	1	4.5
Ill-dim	360	352	98	1	4.5
Ill-bright	360	351	98	1	3.9

There was a significant degree of variation between the six groups (Friedman test: Chi-Square = 26.5, n = 9, p < 0.001).

Pairwise analysis of the six test conditions was performed and the results are summarised in Table 3.

Table 3. Pairwise analysis of the six test conditions for n = 9 test subjects (Wilcoxon Signed Ranks Test).

	Non-dark	Non-dim	Non-bright	Ill-dark	Ill-dim	Ill-bright
Non-dark	-	-2.7 p < 0.01	-2.7 p < 0.01	-2.7 p < 0.01	-2.7 p < 0.01	-2.7 p < 0.01
Non-dim		-	-1.1 n.s.	-2.1 p < 0.05	-2.1 p < 0.05	-1.1 n.s.
Non-bright			-	-0.5 n.s.	-0.5 n.s.	-0.3 n.s.
Ill-dark				-	0 n.s.	-0.4 n.s.
Ill-dim					-	-0.4 n.s.

Discussion

Hypothesis 1: Is there a difference in data input success for partially sighted test subjects versus test subjects with good vision?

For the majority of the testing conditions, the partially sighted test subjects showed a similar degree of data input success as the test subjects with good vision. The only significant difference was in the Ill-bright testing condition, where the illuminated keypad was used under bright ambient lighting. Under these conditions, the partially sighted test subjects demonstrated a small but significantly lower success level. This result suggests that, under bright ambient lighting conditions, the internally illuminated keypad labels might be slightly harder to read.

Therefore, in the case of 'bright' ambient lighting, the null hypothesis (that there is no significant difference in the data input success for partially sighted test subject versus test subjects with good vision) can be rejected.

Hypothesis 2: Does internal illumination on a keypad improve data input success?

The data input success was significantly higher for the internally illuminated keypad than for the non-illuminated keypad. This result suggests that there is a benefit to internal illumination (but see Hypothesis 5 below for a more detailed breakdown of the significant results). Therefore, the null hypothesis (that there is no significant difference in data input success between the non-illuminated keypad and the internally illuminated keypad) can be rejected.

Hypothesis 3: Does ambient lighting affect data input?

Under 'dark' ambient lighting the internally illuminated keypad was significantly more successful than the non-illuminated keypad. This result suggests that the benefit of the internal illumination is most notable under lower ambient lighting conditions. The null hypothesis (that there is no significant difference in data input success under 'dark', 'dim' and 'bright' and ambient lighting conditions) can therefore be rejected.

Hypothesis 4: Does internal illumination level have an effect on data input success?

There was a small but significantly higher degree of success with the 'medium' and 'high' internal illumination levels, when compared to the 'low' level. This suggests that, when providing internal illumination, in order to elicit the maximum benefit from this feature, between 4.5V and 6V would be preferable to 3V. The null hypothesis (that there is no significant difference in the data input success for the three different internal illumination levels) can therefore be rejected.

Hypothesis 5: Is there a difference in data input success on an internally illuminated keypad compared to a non-illuminated keypad, under different ambient lighting conditions?

The internally illuminated keypad was significantly more successful than the non-illuminated keypad under 'dark' and 'dim' ambient lighting conditions. There was no significant difference between the two keypads under 'bright' ambient lighting conditions.

The non-illuminated keypad under 'dark' ambient lighting conditions was considerably (and significantly) less successful than the other five test conditions.

The null hypothesis (that there is no significant difference in the data input success for the two different keypads under different ambient lighting conditions) can therefore be rejected.

Test Subject Feedback

Because of the design of the data input software, it was not possible to accurately test the function keys, however test subjects were asked to comment on the keys and feedback was recorded.

With regard to the "numeric" keys, in bright ambient lighting, four of the nine test subjects preferred the non-illuminated keypad, one preferred the internally illuminated keypad, and the remainder (n = 4) had no preference. In dark ambient lighting, all of the test subjects preferred the internally illuminated keypad.

With regard to the function keys, the users found the labels on the internally illuminated function keys easier to read. Test subjects had a particular problem reading the black on red in the non-illuminated.

One test subject with severe colour blindness found the Cancel and Clear keys on the internally illuminated keypad difficult to distinguish. He commented "the closer a colour is to a pure primary colour, the easer it is to distinguish from other colours".

Final Conclusions and Recommendations

In conclusion, the results of this study suggest that there is a benefit to internal illumination in a keypad in lower ambient lighting conditions. In contrast, in bright ambient lighting, there is a slight disadvantage to having internal illumination, when compared to a non-illuminated keypad with white text on black keys. Figs. 2a, 2b and 2c demonstrate quite effectively how the benefit of internal illumination decreases as the ambient lighting level increases from dark to dim to bright.

On the basis of these results, internal illumination can be recommended for lower ambient light conditions, such as a shaded cash machine, an indoor ticket machine, an underground car park, or a covered petrol station, where artificial lighting may be low or non-existent. The darker the ambient lighting, the more beneficial internal illumination will be.

With regard to level of internal illumination, three levels were tested in this study 3V, 4.5V and 6V. Based on the results of this investigation, an illumination level of between 4.5V and 6V is recommended.

It should be noted that Glaucoma can cause the eyes to be very sensitive to light and glare, and medications can worsen this problem. A partially sighted volunteer with Glaucoma was asked to test the keypads and he was unable to read the keys at any of the internal illumination levels. Glaucoma is found in approximately 2 per cent of the population over the age of 40. It is estimated that more than 500,000 people suffer from Glaucoma in England and Wales alone, with more than 70 million people affected across the world. In 2002 Glaucoma accounted for 12.3% of blind people globally (Resnikoff et al. 2004).

References

Resnikoff S., Pascolini D., Etya'ale D., Kocur I., Pararajasegaram R., Pokharel G.P. and Mariotti S.P. (2004). Global data on visual impairment in the year 2002. Bulletin of the World Health Organisation, 82: 844-851.

Acknowledgements

This project was funded by the RNIB. The author wishes to thank the anonymous test subjects who took part in this study. Sincere gratitude is extended to the staff of RNIB Bristol, in particular David LePoidevin, for considerable help. Thank you to Storm Interface for providing the keypads.