A new font for digital television subtitles
By Janet Silver, John Gill, Christopher Sharville, James Slater and Michael Martin
[Fig2: dot matrix letter 'A']
The basic characters look fairly coarse, especially when the elements include diagonal lines ('font' is really too kind a word for this type of dot-matrix display!) and so 'character rounding' can be used to improve the resolution. Character rounding modifies the video signal from the character generator when a diagonal line is detected, inserting a series of appropriately placed half dots on the interlaced line to smooth out the characters. Effectively, putting different information on the two adjacent interlaced lines doubles the character resolution from 7x5 to 14x10, making for much smoother-looking characters. If a non-interlaced display is used to reduce flicker, as described above, character rounding cannot be used.
Although the basic alphanumeric information for Teletext is transmitted as a 7x5 dot matrix, most modern receivers use a character generator ROM which can provide a 9x10 matrix display, in order to allow for character ascenders and descenders to be more attractively displayed.
In current television systems the ITU-R international standard Recommendation 500-3 recommends that for critical viewing the viewer should sit between four and six times the screen height (i.e 4H-6H) from the television screen. These figures are intended to apply to professional viewers who are assessing the technical quality of pictures. A survey in the early 1980s showed that the average domestic viewer actually watched from a distance of more than 7H. If we make the reasonable assumption that our room sizes are unlikely to change in the short term, then if larger screens are introduced into our living rooms we will be unable to avoid effectively sitting closer, in terms of screen-heights, to the screen. Work has been done relating to future large screen HDTV displays which shows that with the long sought one-metre diagonal flat screen in the home, the most pleasing results are obtained when the viewer sits at 3H from the screen. Interestingly enough, film enthusiasts have long known that the best seats in the cinema are those located about 3H from the screen!
[Fig.3: Comparison of viewing distances for conventional and large screen TVs]
Since it seems likely that larger screen displays will find their way into our homes, it would therefore make sense for a font designed for modern TV use to remain acceptable when used with these larger displays.
Before considering typeface design for any purpose, it is important to understand the fundamentals of the vision process and the behaviour of the normal eye when reading: For the purposes of this paper, it is most useful to consider visual acuity in terms of the angle subtended at the retina: knowing that the normal eye can differentiate a character subtending 5 minutes of arc. With this established it is quite simple to create an algorithm to define for any distance what a person with normal vision, and for that matter what a person with reduced vision, should be able to read as sub-titling on a television screen. For this paper we are defining normal vision as the standard required to drive in the United Kingdom: most authorities agree that this is the equivalent of 6/9 (Charman. 1997).
In the UK about one million people have a registerable visual impairment (they are "unable to do any work for which eyesight is essential" with a recommended level of 3/60 or "are significantly and permanently handicapped" with a recommended level of 6/24) (Bruce et al, 1991). It should be added that screens attract dust which reduces contrast, many people do not have an up-to-date spectacle correction which in any case was prescribed for optimum vision at 6 metres, and the lenses may be damaged or dirty even if they are correct.
At a very early stage in the process of considering the legibility of subtitles, a decision was made by the team to return to basic principles of visual perception and establish a list of criteria before looking at what was available. Any research must start with a survey of work published by others: although a great deal has been written about print, most of it seems to be rather old and consequently quite difficult to obtain. Much of the more modern work appears to address the problems of reading disability (or dyslexia) rather than normal reading ability with normal vision or visual disability. A search through available fonts disclosed many that satisfied most of the criteria but none that satisfied all of them. Therefore a font had to be designed to be as clear as possible while being different enough to satisfy copyright restrictions.
Background and Contrast
Reading continuous text is a rather different task from recognising symbols: the latter task does not have the benefit of contextual information, nor does it require speed or intellectual processing. The eye does not move across print in a simple straightforward flowing movement; rather it moves in a series of saccades (Moses, 1970). The process of vision recognises the edge between contrasting figures against their background. Normally the eye can recognise fairly low levels of contrast, but as a consequence of both age and disease contrast sensitivity reduces. It follows that the greater the contrast the easier it will be to read the material. It is also important to appreciate that the photo-receptors in the retina are stimulated only by light. Tinker (1963) reported on studies, mainly in the 1920's, that demonstrated a clear preference for black ink-print on a white ground, but in later studies (Mehr et al 1973, Ehrlich 1987) using CRT displays the preference was reversed. We are so familiar with dark ink-print on a light background that we tend to consider that to be optimal. In fact the later findings are more compatible with the physiological theory, the eye does not see the black letters, it is stimulated by the white background. The alternative of white text against a dark background produces a stimulus only by the characters themselves. The difference may not be significant in many cases, but it certainly will be if there are any defects in the optical media of the eye, or scatter produced by fine scratches on the surface of spectacle lenses. The resultant noise' in the system may produce discomfort and reduced visual performance. The same is true for anyone with light sensitivity. Recent research using CRT displays has shown that white on black is preferred by the largest number of people, with white on dark blue being the second choice (Silver et al 1995). Quite possibly other preferences are aesthetic except in rare pathological cases.
There have been many suggestions made as to the most visible contrast alternatives, most often based on very dubious theory. For example some years ago there was a fashion for yellow tinted lenses for night driving: the argument being that yellow light is focused exactly on the retina whereas red is focused slightly in front of it, and blue light behind it. Although this is absolutely correct it assumes that the light is coming from infinity, and the observer's eye has absolutely no refractive error - which can seldom if ever be the case and is certainly quite irrelevant to domestic TV viewing which usually takes place in a room with fairly low general illumination.
Other factors contributing to legibility must be considered too. Shaw (1969) in a study of 288 adult and 48 partially sighted people, found that size is the most important factor, but the density or weight of the print is significant too, and it was far more difficult to discriminate letters that are closely crowded together. Such factors are self-evident to optometrists or ophthalmologists, but fonts are normally designed by graphic artists who are primarily interested in aesthetic criteria. It is obvious that the more different each letter is from those around it the fewer errors will occur. Decorated letters will fuse and create noise in the system. Script is considerably more difficult than standard print. Aesthetic qualities, while theoretically not terribly important, will actually have a considerable effect on any individual's desire to read for long periods.
The project was limited to white characters on a black background. However bold and italic are used for emphasis and other colours are used in subtitling to indicate a change of speaker, but time constraints in the first phase of this project did not permit the consideration of these variations.
For the actual design of the subtitle typeface we considered the following aspects:
1. The weight of the characters, i.e. the thickness of the character strokes. If the weights are too thin there will not be sufficient contrast. If the weights are too thick the 'counters', the shapes inside characters such as a e g or d', will be too small and on TV screens these will tend to fill in, thus blurring the centres of the shapes.
2. The width of the characters, i.e. whether it was better to have wide or narrow characters. Current TV subtitling uses the Alphamosaic subtitling font which has a narrow character shape, this enables the type to be displayed in a fairly large size. Helvetica is also used for some subtitling, this is not as narrow as the mosaic typeface. However, sometimes Helvetica is used with spacing so close that the characters almost touch each other, which reduces legibility.
3. The lower case 'x' height of the characters in relation to the ascenders and descenders.
4. The simplicity of the shapes and what was needed to differentiate one character from another when there was the possibility of confusion as in 8 and 3 for example.
5. The shapes of the dots above characters and the shapes of punctuation.
6. The spacing between characters and the spacing between lines of type.
Working together and contributing our expertise as necessary, we designed a typeface with the following attributes:
1. A medium weight that was neither too bold or too light. This enabled the counters of characters such as the lower case 'a', 'e', 'g' and 'd', capital 'R', 'B' and 'P' and numerals '6', '8' and '9', etc. to be as large as possible.
2. A slightly condensed width to the characters.
3. A fairly high lower case 'x' height ratio with ascenders and descenders that were not too long.
4. All of the letter shapes were kept simple. Decisions were made on such things as how the cross bar on the capital 'Q' would fit; how the numeral one' could be distinguished from the lower case ell'; how far to bring round the curves on the capital 'C', 'G' and 'S' and numerals such as '2', '6' and '9'.
The main design aim was to arrive at characters that could be distinguished from each other as easily as possible. Sloping the vertical strokes on some characters made a difference, although this was kept to a minimum.
5. Dots and punctuation had to have simple shapes. Subtleties on these would be lost on TV screens.
6. The spacing between characters and words was carefully assessed.
Kerning has also been applied to the font. This is the specific subtle adjustment of spacing between character pairs, i.e. when a capital 'A' sits next to a capital 'T' (AT) the kerning is adjusted differently to when a 'N' sits next to a 'T' (NT).
Within the available time proper test procedures using stratified samples and a double blind design were simply not possible, but some measure of the response of key viewers was desirable. We were primarily interested in two groups of viewers, the visually impaired, and the hearing impaired.
1. Visually impaired subjects
For visually impaired people a simple test procedure was developed: A standard sentence was printed in three fonts 1. Standard AlphaMosaic , 2. Tiresias (first version); and 3. Times New Roman (this was selected because it is the most widely used font and therefore likely to get the familiarity' vote). All were actually printed in 14 point. To simulate the subtitling situation the examples were viewed in the clinic through a closed circuit television reading machine which enlarged them if needed and reversed the polarity providing well contrasted white text on a black background. Patients attending one of the researchers' Low Vision clinic who mentioned difficulty with television were shown the examples in random order and asked to decide which was the easiest, which the most difficult to read, and the reasons for their choice. Many patients attend with an escort and some of these who expressed an interest were used as controls and to expand the numbers. People were excluded if they were not competent to understand or co-operate.
A later version of the Tiresias Screenfont typeface was produced with improvements to the kerning after the testing commenced; the test sentences were not revised to maintain coherence.
2. Hearing impaired subjects
A short video was prepared using the new font in four alternative sizes: A (30 lines); B (20 lines); C (24 lines) and D (26 lines). A programme that had actually been transmitted was used with only the font changed. The subtitles appeared in white on a black strap just above the bottom of the screen. Groups of hearing impaired people from organisations for deaf and hearing impaired people were invited to view the video under controlled conditions and complete a short questionnaire. A small group of people with normal hearing plus escorts, signers etc acted as controls.
In all 35 visually impaired subjects were used, 9 males and 26 females. The age of the subjects averaged 60.4 years (8 years - 95 years). A number had more than one disability:- several were wearing hearing aids and four needed assistance to move about. Many of the patients had complex aetiology of vision loss, and in such cases the disorder most likely to cause reading difficulty has been selected as the primary diagnosis. The most common diagnoses were one form or another of macular disease (14/26), while this may reflect other factors, diseases of the macula are recognised as being responsible for around half of all registration of visual disability at the present time (Bird 1998). This proportion likely to increase as the population ages. Although the Tiresias Screenfont typeface was the first choice for the majority of the subjects (17/26), 8 preferred Times New Roman. The reasons given were interesting:- in nearly every case the subjects selecting the Tiresias typeface described it as "clearest" or "easiest to read", while those selecting Times New Roman commented on it's "elegance" and "better spacing". AlphaMosaic was immediately identified as worst' by all except one of the subjects, although many hesitated over the choice of best'. The one subject who preferred AlphaMosaic gave "thickest" as the reason. There does not appear to be a significant difference between the controls and the patients, although the sample is too small for certainty; this would support our general philosophy that good design is of benefit to all.
Because problems with subtitles is a task often mentioned by low vision patients, the subjects were also asked the size of their screen, and how far from it they normally sit. Many of the subjects sit quite close to the screen of their domestic television set, but the distance tends to be a function of the furniture arrangement rather than optimum vision, many seem to sit rather farther away than the 'ideal' distance. Many were very vague about the size of their screen, so this data is not quoted.
2. Hearing Impaired People
The video was shown to a total of 62 people. Of these 14 could be classified as controls.
The group consisted of 48 hearing impaired people, divided equally between the sexes, the average age was 62 years (17 to 94, although this average is skewed by 5 people under 21). The responses to the questions can be seen in the attached copy of the questionnaire (Table1).
It proved extremely difficult to direct peoples' attention only to the font, many of those who claimed to prefer the present font did so on the basis that speakers are distinguished by colour or position on the screen in some types of subtitling. While some subjects were keen to have colour, others complained of problems with certain colours though the 'problem' colours differed, many people commented that the white on black was easier to read than colour. There was a widely held opinion that the strap was wider than it might be, too far up the screen, and actually covered the mouths of some of the speakers. This remark was made at all levels of hearing impairment.
A number of people remark that they were very used to the current subtitling format, that this had all the advantages of familiarity, a new font would take some getting used to. Typical positive comments: "much appreciated in comparison"; "much clearer than the usual colours"; "just right"; "a big improvement"; "jolly good", etc.
Of the people with normal hearing there was no greater agreement: 9/14 used sub-titles "sometimes", none always. The size preferences were: A=2, B=3, C+5 and D=3, 12/14 could read the smallest text comfortably, and 10/13 considered it easier to read ( 1=harder, 2= the same, one did not express a view). They were all happy with the spacing, but 11/14 felt that the largest text obscured too much screen.
We recommend that the strap should be no larger than necessary, and then only sufficient to provide a background for the text, and can be supported as sufficient for discrimination so long as there is at least the width of one element of a character in black between the edges that character and the background .
Name/ Sex/ Year of Birth
Harder to read ? Easier? About the same?
It is interesting to note that few, if any, of our subjects in either group are sitting at the 'ideal' range. For most screens this would be less than two metres, and the data suggest that many people over 60 would need a special spectacle correction for optimum viewing.
It is freely conceded that this 'testing' is far from ideal and could even be described as anecdotal: all the subjects were to some extent self selected, and they were in no way stratified or subjected to any of the research criteria normally adopted by the writers. However, it was generally agreed by the DigTT Group that the opinion of potential consumers was essential, and no more could be attempted within the time-scale. It was hoped that comments would indicate areas where more work needed to be done. Although many helpful comments were made, some of the improvements suggested had already been made after observation by the team, and others were outside our remit. In fact the criteria were so well satisfied, and the general approval so consistent that the team is confident that the new font, at the very least, represents a considerable improvement.
There have been criticisms: the subtitles in the video were held to be "too conspicuous" by one professional observer, another felt that it would be "irritating to read in large blocks". However, the font has been designed only against the criteria for subtitles, and may be improved in the light of experience and further constructive criticism. The issues of alternative sizes, colour contrasts, emphasis, other orthographies, graphics, etc must also be addressed in the near future.
The manufacturers of the set-top boxes are pressing to start production, and font chips must be incorporated. Considering the time scale within which we have had to work, this preliminary testing was the best that could be managed. It does at least indicate that the new font is well accepted by our key groups, but it can be considered no more than pilot study.
Further Research Needed.
It has been suggested that the Tiresias font may have wider applications than use for subtitles: but ink-print, public display systems, computer screens etc., present environments with different constraints and these must be addressed appropriately.
We are grateful to The Royal National Institute of the Blind for funding this work; Allen Mornington-West who chairs the DigITT Team at ITN and has been a great source of inspiration and encouragement; Gerry Stallard of ITC who ensured that we satisfied their guidelines; Andrew Martin of Bitstream UK. for loading the typeface into a set top box so that we could assess it under realistic technical conditions; Charles McGhie of the BBC for technical advice; Peter Weitzel of the BBC who organised the production of the video; Garry Duguid of ITFC for light meters and technical support on testing; Ruth Myers, Christopher Meyer and Penny Beschizza for organising subjects; and to all members of the DigITT Team for their constructive criticism and support.
Bird, A. C. (1997) Personal communication.
Bruce, I., McKennel, A. & Walker, E. (1991) Blind and partially sighted adults in Britain: the RNIB survey. London.
Ehrlich, D. A. (1987) Comparative study in the use of closed-circuit television reading machines and optical aids by patients with retinitis pigmentosa and maculopathy. Ophthalmology and Physiological Optics, 17(3), 293-302.
Garzia, R. P. (1996) Vision and reading. St Louis: Mosby.
Gill, J. M. (1997) Access prohibited? information for designers of public access terminals.
Independent Television Commission (1999) ITC guidance on standards for subtitling. [accessed 04/10/07].
Mehr, E. B., Frost, A. B. & Apple, L. E. (1973) Experience with closed circuit television in the blind rehabilitation program of the Veterans Administration. American Journal of Optometry, 50(6), 458-469, June.
Moses, R. A. (1970) Adler's physiology of the eye (5th ed.). St Louis: C V Mosby Co
Shaw, A. (1969) Print for partial sight: a research report. London: The Library Association.
Silver, J. H., Gill, J. M. & Wolffson, J. S. W. (1995) Text display preferences on self-service terminals by visually disabled people. Optometry Today, 35(2), 24-27, 30 January.
Stallard, G. (1997) Personal communication.
Tinker, M. A. (1963) Legibility of Print. Iowa: Iowa State University Press.
Janet Silver OBE., DSc., M.Phil.,FBCO., was Principal Optometrist at Moorfields Eye Hospital, and is now Low Vision Consultant at the John Saunders Suite, Moorfields Eye Hospital. She has concentrated on the aspects of the design to optimise discriminability for people with low vision.
John Gill PhD.,C.Eng.,MIEE., is Chief Scientist at Royal National Institute of the Blind. His role in this project was project leader/coordinator.
Jim Slater is Principal Consultant with Slater Electronic Consultants. His contribution is on relevant technical aspects of digital television.
Chris Sharville is Design Consultant at Laker Sharville Design Associates. He and his staff have been responsible for the design aspects and converting the typeface to a digital format.
Mike Martin was formerly Chief Scientist at Royal National Institute for Deaf People. He provided valuable insights into the behaviour and needs of hearing-impaired people, and did some testing using the video;
Given knowledge of the individual's visual acuity and screen size it is quite possible to calculate the greatest distance at which any individual might sit and still be able to read the subtitles. If the screen height is 227mm, each capital letter will have a height of approximately 10mm which suggests a maximum viewing distance of 4.3 metres for a person with visual acuity of 6/9. However, this represents a threshold, and few people would read rapidly or with comfort for long periods at this range. Working within the thresholds, tables have been constructed using four common screen sizes and suggesting the maximum distance from which the subtitles would be legible for people with visual acuity ranging from the normal (6/9) to well within the legal definition of 'blind' (see Table 2). It should be emphasized that these are rough calculations, and have been rounded for simplicity. Inevitably they will not apply to every single individual - there are far too many variables - but they should hold true for the vast majority. If the criteria change, modified tables can be constructed very easily.
Table 2 Viewing Distances as a Function of Visual Acuity