Skip to Content

AFBAmerican Foundation®
for the Blind

Expanding possibilities for people with vision loss

Five Tips for Teaching Speech Recognition to People with a Visual or Physical Impairment

By Luke Scriven

Over the last two years in my position as an Assistive Technology Specialist at The Chicago Lighthouse, I have trained a number of people who are blind or have low vision on how to use a computer. Our trainees' level of prior computer knowledge has generally been very low. This is mainly due to the age range of the clientele, who tend to be seniors looking for a way to ease into the communication age. Many of them have lost their vision fairly quickly, and have no keyboard skills.

Why Speech Recognition Can Be Helpful for Computer Users Who Are Blind or Visually Impaired

Although learning keyboard layout is productive and a valuable skill, for many users this can be a daunting task, especially when combined with learning to use the software. In addition, many of our clients also have physical disabilities which make it difficult for them to use a keyboard or mouse. In these situations I turn to speech recognition—in particular, the software program CDesk from AdaptiveVoice—to help bridge the gap between what they want to do and how they can achieve that goal. Note that there are a number of good solutions for speech recognition such as Dragon Naturally Speaking and Microsoft's own speech recognition service, as well as mobile solutions such as SIRI on the iPhone.

Tips for Helping Visually Impaired Seniors Learn to Use Speech Recognition Software

Here are some general pointers that I have found help visually impaired seniors learn how to use speech recognition:

1) Adjust the microphone levels: The level must be loud enough for the user's voice to be heard clearly, but quiet enough to avoid voice clipping. (Voice clipping: when the input level from the microphone exceeds the maximum level allowed by the software, the extra signal which is beyond the software's capability is simply cut off, which causes distortion of the input sound.) The Microsoft speech recognition control panel, visible whenever speech recognition is on, provides useful feedback in the form of a VU meter on the right which shows the level of input and turns red at the top whenever voice clipping is occurring.

For some users who have a weak or variable voice pattern, it may be useful to create more than one speech profile. One profile can be used for the start of the day when their voice is stronger, and one for later when their voice is more fatigued. In CDesk, users are able to swap between profiles through an accessible menu. It is important to set appropriate volume levels before beginning any speech training.

2) Mute the microphone when not talking to the computer: This is especially a problem for users who are visually impaired as they may cause something unexpected to happen and then not be sure what has happened and how to get out of it. The need to mute the microphone may have to be reinforced multiple times for those with memory difficulties.

3) Make students aware that you cannot talk to the computer like another human being: It is natural for someone unfamiliar with the technology to assume that you can talk to the computer as you would another person. It's important to manage expectations in this regard to ensure the correct voice commands are being used to perform the necessary functions.

Command and control are likely to work well immediately. For example, in CDesk the 'Peggy' commands such as 'Peggy, start Word,' 'Peggy, exit,' and 'Peggy, what time is it?' will work for the vast majority of users without any training and almost irrespective of the quality of their speech. Command and control are easier for the computer than dictation because it is listening for a very specific set of words. This is opposed to dictation, where the user could be saying anything.

4) Repetition is key! Repetition is the key to success, particularly with users who have difficulty with memorizing commands.

5) Tackle dictation in shorter phrases, rather than a 'stream of consciousness' style: This approach allows users to check what is being typed as they go, making it easier to make corrections when necessary. For users with certain physical impairments speaking in shorter phrases is especially important, as they run out of air more quickly, changing the timbre of the voice and leading to incorrect speech recognition.

In my experience, failure is very unlikely to be caused by a user's accent or speech impairment, although the quality of recognition may be lower initially, particularly for dictation. This is usually remedied over time as the computer builds a profile of the user.

Adaptations for Users with Physical Impairments

For clients who have a physical impairment, there are other interesting ways to use assistive devices.

1) One example of a device that can be used to aid with speech recognition is a Bluetooth headset. This allows users with physical impairments to have the microphone accessible without the need to physically take a headset on and off whenever they want to change positions.

The main limitation of Bluetooth headsets is the voice fidelity, which is much lower than, for example, a USB-corded headset. This leads to a lower quality for audio both received and transmitted through the headset. This can be problematic in speech recognition where the quality of signal reaching the computer from the microphone needs to be as clear as possible in order to achieve a good percentage of accuracy.

The other issue with a Bluetooth headset is battery life—typically around 6 hours when in use, longer when in standby. In order to go into standby mode, however, the user needs to physically interact with the headset. Despite these limitations, in some cases the benefits of a Bluetooth headset may outweigh the cons.

2) Another innovative solution tried by a colleague of mine is the use of a monitor stand clamped to the front of a desk, rather than the back, to assist users who are confined to a wheelchair. The articulating arm of the stand is positioned to come out from the front of the desk, meaning the user can more easily position their wheelchair in front of the monitor, without the desk being in the way. The screen can also be positioned at the appropriate height, allowing the user to get their face very close to the screen, a necessity for some people with low vision.

These solutions use commercially available products and may be of great benefit to the user. The user's physical comfort can be just as important a factor as psychological comfort in ensuring a successful outcome.

Conclusion

In conclusion, speech recognition can be a powerful tool for clients who are unfamiliar with keyboard layout and have a visual or physical impairment, but it can also be frustrating to learn. When you are teaching speech recognition software, particularly to people with a physical or visual impairment or those with memory difficulties, following the five tips discussed earlier can greatly assist in making speech recognition consistent with user expectations.

services icon Directory of Services

book icon Featured Book

JVIB Special Issue on Critical Issues in Visual Impairment & BlindnessJVIB Special Issue on Critical Issues in Visual Impairment & Blindness

JVIB Special Issue on Critical Issues in Visual Impairment & Blindness

Join Our Mission

Help us expand our resources for people with vision loss.