4. Discussion, Conclusions and Guidelines
The aim of this study was to evaluate different user interface options and to identify general issues faced by visually impaired users. A major conclusion that can be drawn from the results is that (with the exception of the ability to read the on-screen instructions) most of the problems encountered by the VI test subjects were also encountered by the test subjects with good vision.
4.1. Problems identified with fingerprint readers
4.1.1. Problems with instructions
The most critical problem observed in these fingerprint reader evaluations was that the majority (96%) of visually impaired users were either unable to see the screen at all, or did not have sufficient sight to read the instructions on the screen. As a result, these users were unable to enroll a fingerprint unaided.
For the purposes of this study, the evaluator read the manufacturer's instructions directly from the screen. When this was done, the number of VI test subjects who had at least one successful fingerprint image taken significantly increased for all three fingerprint readers when compared to the number before the instructions were read out (i.e. 0% for all three readers). Although this result may seem obvious, it highlights that none of the readers were usable by the VI test subjects before the instructions were provided in an accessible format (i.e. audio).
When more detailed instructions were provided the number of VI test subjects who had at least one successful fingerprint image significantly increased again for the IBM and Microsoft readers. This indicates that the manufacturer's instructions were not always sufficient to explain to visually impaired users how to enroll a fingerprint. The modifications made to provide "more detailed instructions" were not substantial, and in many cases involved simplifying the wording or compensating for the fact that the user could not see an accompanying image.
To provide an example, the IBM manufacturer's instructions provided the following:
To use the fingerprint reader, place your fingertip on the small circle above the reader, then apply light pressure and swipe your finger across the reader in one smooth motion.
A majority of the VI test subjects did not understand this explanation, as they could not see the accompanying animation demonstrating how to move the finger. Many guessed that the finger should be swiped from left to right, rather than down the length of the reader, and some commented that a "swipe" indicates a swift action, which would not be appropriate. So the following, more detailed, instructions were provided when needed:
The fingerprint scanner area is the small strip below the indented circle. Put your fingertip on the circle. Moving your finger down the length of the reader, towards your body, slowly stroke your finger over the scanner area, so that your finger is touching the strip at all times as you pass over it.
These instructions were sufficient for all VI test subjects to understand how to correctly interact with the reader. Instructions should therefore be as descriptive as possible, while at the same time keeping the wording simple.
4.1.2. Problems encountered when positioning the finger into the correct area on the reader
Many test subjects in this study commented positively on the fact that the Hamster and Microsoft readers had raised areas surrounding the scanner area itself, which naturally guided the finger into the correct location. This was corroborated by the high success rate on Task 2, where 89% and 96% of VI test subjects correctly guessed the correct area to place their finger on these fingerprint readers.
Although the Microsoft reader had a high success rate overall, many test subjects commented negatively about the large size of the scanner area. This however is, presumably, to accommodate people with larger finger size. As functional as it might be, it caused problems for some test subjects who were enrolling their index finger, as they didn't always have their finger centred correctly. One possible solution to this is to have a non-slip finish on only the centre of the scanner area, leaving the remaining scanner area glass-like (i.e. two different textures). This way the user will still benefit from the non-slip effect, will know when the finger is centred, while also allowing for people with larger finger size.
4.1.3. Problems recognising a success or failure
With regard to the IBM and Hamster readers, when a VI test subject placed a finger onto the reader, the subject did not know whether or not each scan had been successfully taken. The evaluator had to inform them of this.
The Microsoft reader, on the other hand, provided an audible signal (two different combinations of notes indicating a success or failed scan), which encouraged even the more uncertain test subjects.
The audible signal was not, however, noticed by all of the test subjects. Furthermore, the different sounds used to signify a success and failure on the Microsoft reader were not ideal, as they were not sufficiently distinctive. A success should be signified by a higher, more pleasant sound (e.g. chimes ringing). A failure should be indicated by a lower, less pleasant sound (e.g. buzz).
4.1.4 Problems with not understanding why a scan failed
Of those who experienced at least one failed scan, very few of the test subjects understood what they could have done differently to avoid a second failure.
Two of the three readers (IBM and Microsoft) provided a limited amount of interactive information, instructing the user on what he/she was doing wrong: "move right", "move left", "slow down" (IBM), "place [your finger] flat on the fingerprint reader" and "centre [your finger] on the fingerprint reader" (Microsoft). Although some test subjects did comment that, in theory, these instructions were helpful, they did not always help to rectify the problem.
When the test subject experienced a failed scan followed by a successful scan, the evaluator asked "do you know what you did differently on the second scan?" A significant majority of the VI test subjects did not notice that they had done anything differently and could not explain why a scan was unsuccessful (Task 11).
4.2.1. Making the reader easier to locate
- The scanner area should be lit internally. The light should only turn on when the reader is awaiting input or the light should start flashing slowly when the reader is awaiting input.
- Visually the reader should stand out if it is attached to a terminal. The reader could be highlighted by making it a different colour to the surrounding area.
- The reader should not be flush against the rest of the terminal. This will make it more noticeable from both a visual and a tactile point of view. The device as a whole should be raised from the terminal, with the scanner area itself slightly recessed into the raised surrounding casing.
4.2.2. Improving the accessibility of the fingerprint reader design
- A simple action should be required to enroll a fingerprint, such as simply pressing a finger on the scanner area and removing it. This is the action that 98% of the test subjects (including those who took part in either the pilot or the study) assumed they should use.
- Ideally the user should be able to simply press and hold the finger in place, while the multiple scans are taken in sequence (i.e. removing the necessity to lift and replace the finger repeatedly). This would cater for people with, for example, reduced manual dexterity or a hand tremor. However, if the movement of the hand is part of the authentication process (i.e. reducing the chance of the reader being fooled) this may not be an acceptable option.
- The reader should be centred on the terminal, so it is equally accessible for right- and left-handed people.
- Of the 40 test subjects in this study, two (5%) preferred to pick the reader up when they were using it. The remaining test subjects did not have any problem using the three readers flat on the desktop.
- Although the option of picking it up in one hand might be favourable for some, this may mean that the reader could be broken, stolen or simply in a different place every time the user goes to find it. Therefore having it fixed to the terminal would be preferable.
- Having the reader fixed to the terminal, but with an additional option of changing the angle at which it sits, could provide a viable compromise for the latter two problems.
- There should be a raised casing around the scanner area. This would allow the user to hold on to the casing, providing support for the repeated removal and replacement of the finger during the registration stage.
- The scanner area should be recessed into the surrounding raised casing, thereby guiding the finger into the correct location.
- Ideally, a tactile marking (consisting of, for example, a circular area with a different texture) in the very centre of the scanner area would allow accurate positioning of the finger, while at the same time not limiting the size of the scanner area, benefiting people with larger fingers.
4.2.3. Providing instructions in an accessible format
- If the terminal is unmanned, or an assistant is not always available to help the user, audio instructions should be provided, taking the user step-by-step through the enrolment and authentication process.
- Instructions should be provided, explaining any progress made.
For example, if a scan is successful: "This scan was successful, please remove your finger and place it on the reader again."
- Any further instructions explaining what the user is doing wrong would also be very helpful.
For example, if the scan is not successful: "This scan was not successful, please move your finger to the right."
or
"This scan was not successful, please hold your finger still on the reader."
- There should be a clear sound to indicate a success and a failure. A success should be signified by a higher more pleasant sound (e.g. chimes ringing), a failure by a lower less pleasant sound (e.g. buzz).
- The user should be told, before the scanning process starts, if it is necessary for more than one scan to be taken for registration.
- When the first scan has been taken there should be an audible acknowledgement (such as a chime sound) followed by a spoken instruction: "The first scan has been successfully recorded. Please place your finger on the fingerprint reader for the second scan." And so on.
- There should be a clear indication when the registration process is complete. For example an audio message "Your fingerprint has been successfully registered."
- If the terminal is awaiting further information, the instruction should say this. If not, it should indicate that the user has reached the end of the process. For example "The registration process is now complete. Thank you."
- If registration fails, there should be a clear indication that the process will restart. For example "The registration has failed because the four fingerprint images did not match. The process will now restart. To enroll a fingerprint, you must have four successful scans. Please place your finger on the fingerprint reader for the first scan. Hold your finger still on the scanner area until you hear a chime."
4.2.4. Informing the user that the reader is waiting for him/her to take action
- The reader should be lit when it is awaiting input from the user.
- The reader should only light up when it is ready to enroll a fingerprint. When the process is complete the light should turn off.
- An audible message ("Please place your finger on the fingerprint reader") would help users who have insufficient vision to see the reader. See Section 4.2.3 above for more detailed guidelines on spoken instructions.
- A timeout feature on the terminal should not be excessively short, as the user may need an extended period of time to find the reader and to complete the required actions.
- If the user is taking an unusually long period of time to respond to an instruction the instruction should be repeated at least once before the terminal times out.
4.2.5. Catering for users who do not require audio instructions (e.g. those who have good vision, or those who are familiar with the process)
- An option to bypass the audio instructions should be provided. This could simply be that the audio comment is skipped or cuts out if the user provides the correct input.
This project was funded by the Royal National Institute of the Blind. Thank you to John Gill and Julian Jones for comments. The author wishes to thank the 45 anonymous test subjects who took part in this study. Sincere gratitude is extended to the staff of RNIB Bristol and of Modern Eyes Training Services Ltd. in Gloucester, in particular David LePoidevin, Carole Bennett and Sue Allard for their considerable help, and to Linda Newson of the Scientific Research Unit for extensive background research into Biometric Systems.
Abry, C., Vilain, A. & Schwartz, J.-L. (2004). Vocalize to localize? A call for better cross talk between auditory and visual communication systems researchers. Interaction Studies, 5: 313-325.
Jain, A., Hong, L., Pankanti, S. & Bolle, R. (1997). An Identity-Authentication System Using Fingerprints. Proceedings of the IEEE, 85: 1365-1389.
Jain, A., Pankanti, S., Prabhakar, S., Hong, L. & Ross, A. (2004). Biometrics: A Grand Challenge. Proceedings of the 17th International Conference on Pattern Recognition (ICPR '04).
Lee, H.C. & Gaesslen, R.E. (1991). Advances in Fingerprint Technology. Elsevier, New York.
Miller, B. (1994). Vital Signs of Identity. IEEE Spectrum, 31: 22-30.
Shen, W. & Khanna, R. (1997). Prolog to An Identity-Authentication System Using Fingerprints. An Introduction to the paper by Jain, Hong, Pankanti and Bolle. Proceedings of the IEEE, 85: 1364.
Shen, W. & Tan, T. (1999). Automated biometrics-based personal identification. Proceedings of the National Academy of Sciences, USA, 96: 11065-11066.
Van der Putte, T. & Keuning, J. (2000). Biometrical fingerprint recognition: don't get your fingers burned. IFIP TC8/WG8.8 Fourth Working Conference on Smart Card Research and Advanced Applications, pp. 289-303. Kluwer Academic Publishers.
| < Contents page |