A new teletext character set with enhanced legibility
Citation for published version (APA):
Nes, van, F. L. (1986). A new teletext character set with enhanced legibility. Proceedings of the SID, 27(3), 239-242.
Document status and date: Published: 01/01/1986
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
providing details and we will investigate your claim.
' L. '2
jl~.
,
..
A New Teletext Charncter Set with Enhanced
Legibility
FLORIS L. vAN NES
Abstract-Teletext is difficult to read, partly because of the letter fonts employed. Present fonts are contained in a matrix of 6 (horizon· tal) x 10 (vertical) elements. Research on matrix characters of opti-mum legibillty started in 1969 at the lnstitute for Perception Research. Criteria resulting from this research have now been used to design al-phanumeric characters in a matrix of 12 x 10 elements for use in Teletext. Several verslons of each character were designed and their legibility tested in recognltion experiments. The legibility of the best new version for each letter was compared with and shown generally to be greater than that of the presently used version.
l. INTRODUCTION
D
OT-MATRIX characters are used for text display on conventional TV receivers in an increasing number of consumer-electronics app1ications, such as Teletext, videotex, electronk games, and personal computing. The resolution of a TV display is rather limited because of bandwidth limitations of the TV channel and the video amplifier, etc. Therefore, the dot matrix of which the characters are composed is relatively coarse, implying that they can only be schematic approximations of the elabo-rate detailed fonts used in print. In order to ensure good legibility of such schematic letters and digits, they should be designed with three criteria in mind [ 1]: acceptability, identifiability, and discriminability. A character has high acceptability when its shape closely corresponds to a con-cept that observers have of this shape; it is highly iden-tifiable when its parts stand out clearly against the ebar-aeter background; it has high discriminability when the chances of it being confused with a similar character are low. Such confusion may occur under difficult observa-tion condiobserva-tions, such as low contrast between character and background or reading from a distance. A luminous contrast that is too low occurs, for instance, when red or blue letters are used on a black background, or yellow letters on a white background. With respect to viewing distance, applications such as Teletext are commonly viewed from the same distance as normal TV programs. Aowever, this di stance is too large for the si ze of Teletext characters, which means that, especially for viewers with a reduced visual acuity, Teletext is inherently difficult to read. The following may illustrate this point: the height of the row of capitalletters on a Snellen chart, which canManuscript received October 7, 1985; revised March 4, 1986. The author is with the lnstitute of Perception Research, IPO, Eindho-ven, The Netherlands.
IEEE Log Number 8608715.
be read by somebody with an average visual acuity, i.e., 1.0 (equivalent to 20/20), subtends 5 min of are. To avoid letters with a simHar contiguration being confused during reading a text, it should be made up of letters that are considerably higher, for example 12 min of are to quote a tigure from a human-factors handhook [2].
For a large-screen TV display with Teletext letters of the regular size, this value corresponds to a viewing dis-tanee of 1.6 m, i.e., less than half of that which is typical for viewing TV. Therefore, it is worth optimizing ebar-aeter discriminability.
We designed alphanumeric characters and punctuation marks on a matrix of 12
x
10 elements (horizontalx
vertical, including gaps between letters and rows). Such a matrix allows more refined as well as more acceptable configurations, compared to the 6
x
10 matrix now mostly in use. The latter format presents minimal possibilities for designing upper- and lower-case letters. The resulting character configurations were judged by viewers as being too square with too thin diagonal strokes. To counteract such etfects, "character rounding" was introduced by adding half dots at the appropriate positions, close to the diagonal strokes [3]. The rounding rules are based on an interlaced scan pattem; however, the use of two inter-laeed fields in one TV frame creates an annoying "line flicker" effect when watching Teletext. Most present-day European TV sets therefore do not interlace in the Tele-text mode thus, unfortunately, obliterating character rounding. In view of this outcome and, on the other hand, developments in the Oerman "Bildschermtext" (view-data) service, a 12x
10 matrix format has been recently adopted as the new videotex matrix standard by the Euro-pean Conference of Posts and Telecommunications Administrations (CEPT}.11. DESIGN AND TESTS OF LOWER-CASE LETTERS
The first phase of this project consisled of designing four configurations for each lower-case character using the results of previous experiments on the acceptability as well as discriminability of another comparable character set as guidelines [I]. The new characters we re designed on a terminal screen by assembling matrix "dots" in a grapb-ical representation of the character matrix that was mag-nified approximately 15 times compared to the normal size. The resulting contiguration could subsequently be observed on a TV screen at normal display size. In this way, a stimulus set of 4 x 26
=
104 characters was ob-0018-9383/86/0800-1222$0 I . 00©
1986 IEEE• • • •
(a) (b) (c) (d) Fig. I. Four configurations for the letter a that were used in the first twoexperiments.
tained. Fig. l(a) to (d) shows the four different versions of the letter a.
These 104 eh araeters we re presented in random order to two groups of 12 subjects each in two experiments. In the tirst experiment the characters were presented foveally for 2 s, on a 25-in color TV set (maximum horizontal screen dimension 53 cm, frame rate 50 Hz) at an abser-vation distance of 8 m. At this distance the character box of 12 x 10 dot-matrix elements, as shown in Fig. 1, sub-tended a viewing angle of 4.5 min of are horizontally and 6.5 min of are vertically. In the second experiment, the characters were presented peripherally for 0.1 s to the left or right (in random order) of a tixation cross that was gen-erated in the center of the screen from the same TV set. In this experiment an observation distance of 4 m was used, and the stimuli were presented at an eccentricity of plus or minus 2 degrees; the character box then subtend-ing a viewsubtend-ing angle of 9 min of are horizontally and 13 min of are vertically.
The viewing distance in the tirst experiment and retina) eccentricity in the second one were chosen so that the average recognition score was around 50 percent. This methad allows a clear separation between characters of high discriminability, which then score considerably higher than 50 percent, and characters of low discrimin-ability. which then score much Iower.
Same of the results of the second experiment, in which each stimulus of the set was presented three times to each subject, are shown in Figs. 2 and 3. Fig. 2 represents a confusion matrix for the worst, i.e., least discriminabie versions (e.g., the a of Fig. l(a)); and Fig. 3 is for the best, i.e .• most discriminabie contigurations (e.g .• the a of Fig. l(d)).
The main diagonals of Figs. 2 and 3 represent the cor-rect recognition scores. A comparison of these two diag-enals clearly shows that the differences between the best and worst versions are not the same for all letters. With respect to the confusion, it appears that the errors are more concentrated in particular cells for the least legible letter versions than for the most legible ones: there are 12 cells with a content of 10 or more in Fig. 2, and only two such cells in Fig. 3.
The tirst experiment had yielded similar results. The correct scores from bath experiments were added for each letter contiguration. Generally, the contiguration with the highest combined score was then taken for the tinal char-acter set. However, if there was only a small difference between the combined scores for two contigurations. ac-ceptability criteria were taken into account to choose the contiguration that 1) corresponded most to the intemal
240 Van Nes/Teletext Character Set with Fnhanced l.ePihilitv
~
f
!
~ 16..
"ii
..
15.i
iii Rec:ogJIIzed . . : •..
c d•
I • 4 1 1 1..
1 12 c 1 18 3 1 d 1 1 21 1 ••
•
1 I 14•
3•
I J•
2 ·-1 I 2 m 1.
0 7 2 1•
•
1 1.
••
8 3 I.
.
.
.
7 1 '!_ 1 • 5 1 2•
WORST VERSlONS I h I I k I M n o p Q r • u ' • ? 2 7 2 1 10 7 2 3 1 1 3 1 5 4 1 3 1 1 3 1 2 . 2 3 1 3 1 1 2 1 1 21 1 3 1 10 1 12 1 3 3 20 2 19 5 1 10 12 1 33 2 10 1 11 1 4 5 13 8 2 9 5 12 7 23 1 5 13 1 2 1 2 3 1 • 4 1 1 2 • 17 • ~ : 2031 1 2 2 1 1 1•
1 3 12 1 1 11 1 18 1 , .•
3 3 1 2 1 • 1 1 1 241 • 1 1 1 3 2 1 5 1 2 5 3 • 14 14 5 5 2 2 2 __ _1_~ •• e 1 z:s 10 1 a te ze 11 , . ze 11 •• ' ' 12 11 et a 13 :s• za a z • 10 11Fig. 2, Confusion matrix for the worst versions of the investigated lower-case letter configurations in a discriminability experiment with periph-eral stimulus presentation.
Recoped as : BEST VERStONS
• t t c d e t g h i J I I . I 0 • • r o I u • • • 1823 1 1 1 1 2 8 : : 23 2 1 1 1 2 4 • r a '? 2 d 2 1 27 1 3 2 2
.
'
5 1 3 4 1 2 1 1 1 2 1 2 1 1 1112 1 5 1 125 25 3 1 8 36 ~ 1~ • 1~12 2 1•
1 2 1 • 2 1 1 1 2 7 23. 2 3 30 15 2 3 1 219123 11 2 1 28 3 31 35 2 • 5 4 2 1 1 21 1222~ ~ • 3 26 1 4 1 • 1 1 4 1 1 • 2 1 1 1 1 2 1 1•
•
•
~ --1~25 16 20 3 • 11 1• 3 e 8 33 11 I 10 8 32 11 12 8 ... 1 22 14 18 21 2 I 4 IeFig. 3. Confusion matrix for the best versions of the investigated lower-case letter configurations in a discriminability experiment with periph-eral stimulus presentation.
representation of the character concemed and 2) titted best in the complete alphabet, in the apinion of a few observ-ers. Such a situation was obtained for the two letter a con-tigurations shown in Fig. l(c) and (d); that from Fig. l(c) was considered to be more acceptable, so it was selected for the tinal set of optimally discriminabie and acceptable characters, named "IPO-Norrnal," and as such appears in Fig. 5.
111. CoMPARATIVE EvALUATION OF LowER-CAsE
LETTERS
In the second phase of the project, the discriminability of the IPO-Norrnal set was compared with that of three other sets in a new experiment. The other sets were:
1) "IPO-Bold," with bold versions of the "IPO-Nor-mal' • lower-case letters;
%correct 100~---. 13 short letters 75 50 2S S deseenders IPO-Normol IPO-Bold
Fig. 4. The recognition scores of ascenders, shon letters, and deseenders in a distance-reading experiment with lower-case letters from four al-phabets: Presently used. i.e., 6 x 10, Eisenbeis. /PO-Normal, and /PO·
Bold. The dashed line refers to the average recognition score for alllower-case Jeuers from all four alphabets: 61 percent.
2) a German set of lower-case letters [4]; and
3) the lower-case letters presently used in most teletext decoders, but without "character rounding." In this experiment, lower-case letters from the four al-phabets were presented centrally (at the same viewing dis-tanee as previously used, i.e., 8 m) in random order; 13 subjects participated. Fig. 4 shows the results of the com-parative experiment, separately for the three types of lower-case letters: ascenders, short letters, and descend-ers. A veraged over all the lower-case letters, the recog-nition score of the "IPO-Normal" letters was 65 percent, that of "IPO-Bold" 63 percent, that ofthe German letters 59 percent, and that of the present set 57 percent.
In judging the practical significanee of these results it should be realized that when the characters of such sets are used for representing nonredundant alphanumeric strings, as may occur in codes of all sorts, the probability that the whole string is correctly recognized equals the product of the recognition probabilities for the symbols that constitute the string. Therefore, a difference in rec-ognition probability of a few percent at the level of single symbols can become quite significant at the level of com-plete codes. Bouwhuis [5] has shown that, in principle, the same multiplication rule holds for the recognition of three-letter words when the recognition probabilities of the component letters are known.
IV. CAPITALS AND NUMERALS
Essentially the same two-phase procedure was used for upper-case letters. For numerals, however, a somewhat different route was followed. Three sets of numerals were designed: one set in which the numerals had the same stroke width as that of the upper- and lower-case letters and two sets of boldface numerals with a larger stroke width. The discriminability of these numerals was tested in an experiment, with the numerals from the three sets as stimuli. The boldface numerals scored as high as the others. 1t was then decided to use boldface nu merals in the final character set because the increased stroke width might facilitate the distinction between numerals and
cap-itals in alphanumeric strings. The discriminability of the boldface numerals with the highest correct recognition scores was tested in a new experiment using only such bold digits as stimuli. Some numerals that had an unsat-isfactorily low recognition score, viz. 5 and 6, were then redesigned, taking account of the particular confusion er-rors of the subjects. The resulting set of numerals was again tested; this time the correct recognition scores were more uniformly distributed among the numerals.
V. DISCUSSION AND CONCLUSIONS
Finally, a complete set of 196 characters-alphanumer-ics, punctuation marks, and supplementary symbols-was obtained on a 12 x 10 dot matrix. The most important characters are shown in Fig. 5. All alphanumeric ebar-aeters of the set have a width of 9 or 10 matrix elements, so the capita! si ze is (9 or 1 0)
x
7 _The character design procedure described may be em-ployed in a variety of other applications. With its empha-sis on discriminability, it is especially suited for the de-sign of characters to be read under poor observation conditions.
Comparisons of the IPO-Normal character set with al-phabets designed in other dot-matrix formats, for instanee the ubiquitous VDT font with a capita! size of 7
x
9, are difficult, at least as far as the respective mutual discrimi-nations are concemed, because small differences in dot configurations may entail substantial differences in rec-ognition and confusion scores. For example: a horizontal displacement of the ascending part of the numeral 6 over a distance of one matrix element in the present experi-ments caused a difference in correct score of more than 30 percent, viz. 47 versus 11 percent for the two different configurations, because the perceptual difference with the other numerals, especially the 4, had been increased con-siderably by the displacement.One feature of the described character set, bold numeral strokes, three elements wide-compared with two for the upper-case letters-is not found in the widely used 7
x
9 fonts.\
·-~~·-~
:!.. ... _T~
...~E
E
m
~'!=::_;:;.: :::::::~~-
•
··--· ... . ... ,.,__
~,-.,····::-..
-... ... •;-: :~:·::··::.-
·-...
"'' ::::=::..T-
... ..
. . .-
. •l;...,, •• ._.. ::7~~-~_:"~'•
-~·.:::.··~~. ~; ·~ "-
···~· ·-··:: ~-7:.-ii ±:::::::::: --··-·~~-' ..._~.Fig. 5. The basic set of IPO-Nonnal 12 x 10 dot-matrix characters (copy-righted). The complete IPO-Nonnal set is now proteeled under the rules of the International Design Registration effected under the Geneva Pro-tocol of 1975.
It facilitates the distinction between numeral-capital pairs such as 5-S, 0-0, 8-B in the IPO-Normal set. In passing, it may be remarked that there appear to be few publisbed research results, if any, on the legibility of lower-case dot-matrix letters, whereas there are at least some on the legibility of upper-case letters and numerals [6], [7].
REFERENCES
[I) J. A. J. Roufs and H. Bouma, "Towards linking perception research and image quality," Proc. S/D, vol. 21, no. 3, pp. 247-270, 1980.
{2) H. P. van Cott and R. G. Kinkade, Eds., Human Engineering Guide
to Equipment Design, revised edition. Washington, DC: Amerkan Institutes for Research, 1972, p. 107.
{3) L Reynolds. "Teletext and viewdata-A new challenge for the de-signer," lnformation Design J., vol. I, pp. 2-14, 1979.
[4) M. Eisenbeis, "Visual design ofinfonnation systems," Displays. pp.
95-99, July 1980.
[5] D. G. Bouwhuis, "Visual recognition of words," Ph.D. dissertation, Nijmegen University, 1979.
242
[6) H. F. Huddleston, "A comparison oftwo 7 x 9 matrix alphanumeric designs for TV displays," Appl. Ergonomics, vol. 5, no. 2, pp.
81-83, 1974.
(7] H. L. Snyder and M. E. Maddox, "On the image quality of dot-matrix
displays." Proc. S/D, vol. 21. no. I, pp. 3-7, 1980.
*
Floris L. van Nes received the M.S. degree in electronic engineering from Delft University of Technology, The Netherlands. in 1961, and the Ph.D. degree in physics and rnathematics from the University of Utrecht, The Netherlands, in 1968. Currently, he is working as a research scientist at the Institute for Perception Research-IPO, Eindhoven, where he is the coordinator of all ac-tivities in infonnation ergonomics. His research is centered on the interaction of computers with nonexpert users.
