|
| |
Microphone Usability
Microphones are necessary for speech recognition to function,
and the quality/ergonomics of a microphone can make the different between an unusable
speech product, and one where speech recognition shines. A good microphone will have half
the speech-recognition error rate of a poor microphone.
However, microphones are one of the largest impediments to people
using speech recognition. In general, the failure points are:
- Users don't have a microphone.
- Users don't have the microphone plugged into the computer.
- The sound card isn't compatible with the microphone.
- If the microphone gain is so high that clipping happens then accuracy
will be terrible. If it's too low then accuracy won't be as good as possible.
- Users don't know how to hold/position a microphone.
- Cheap (poor quality) microphones are the norm. (Luckily, this is
starting to change.)
- The most accurate microphones require that the users wears them. Many
users don't like wearing microphones.
If a speech aware application doesn't take all of these factors into
account, many users will be unable to get speech recognition to function well. To avoid
microphone problems, we recommend that your application use the Microphone Setup Wizard (programmatically
available from the Speech Tools API).
Usability of specific types of microphones
If you are considering bundling a microphone with your application,
sound card, or PC, here is some usability information on specific types of microphones.
Close-talk
Benefits:
 | Acceptable close-talk microphones now sell for about $50 (US) at
computer stores, and are worth the expense if users wants maximum accuracy. |
| Highest accuracy of any type of microphone because the element is
close to the mouth, and almost always near-field. |
| Near-field element so it's noise resistant. |
| User has his/her hands free. |
| User can lean back in his/her chair and talk. (Make sure the
microphone cord is long enough so users can lean back, even if their PC is under their
desk.) |
Problems:
| Most users dislike wearing something or being attached to their PC.
Sometimes the cord gets tugged on when the user moves around. |
| It's difficult to talk on the phone with a telephone handset while
wearing the close-talk, unless the phone is routed through the headset. Some headsets have
this functionality. |
| Slow to put on and take off. |
| The user has to reposition the element every time they put the
microphone on. When the element position changes so does the sound of the audio. Users
must be aware of this so they don't make this mistake. |
| Sometimes users position the element directly in front of their
mouth, causing recognition problems when they exhale onto the element. Users must be aware
of this so they don't make this mistake. |
Ear Piece
Benefits:
| Second only to close-talk and handset microphones in accuracy because
the element is close to the mouth. |
| Some ear-piece microphones have a near-field element so it's noise
resistant. |
| User has his/her hands free. |
| User can lean back in his/her chair and talk. (Make sure the
microphone cord is long enough so users can lean back, even if their PC is under their
desk.) |
Problems:
| Most users dislike wearing something or being attached to their PC.
Sometimes the cord gets tugged on when the user moves around. |
| Difficult to talk on the phone with a telephone handset while wearing
the close-talk. |
| Slow to put on and take off. |
Handset
Benefits:
| Second only to close-talk and ear-piece microphones in accuracy
because the element is close to the mouth. |
| Some handset microphones have a near-field element so it's noise
resistant. |
| Users don't like talking to computers, and don't like it when
colleagues walk by their office and see the user "talking to themselves." With a
handset, users feel like they're talking to a person (or at least an answering machine),
and people walking by just see someone talking on the phone. |
| Some handsets have switches that allow the user to turn speech
recognition on/off. |
| Of all the microphones, users are already trained to use this one.
|
| Easy to pick up and put down. |
| User can lean back in his/her chair and talk. (Make sure the
microphone cord is long enough so users can lean back, even if their PC is under their
desk.) |
Problems:
| Users have to hold the handset, which can get tiring if it's used for
long periods of time. |
Clip-On
Benefits:
| Moderate accuracy because it's less than 30 cm from the mouth.
|
| User's hands freed. |
| The microphone element is out of the way. |
| User can lean back in his/her chair and talk. |
Problems:
| Most users dislike wearing something or being attached to their PC.
Sometimes the cord gets tugged on when the user moves around. |
| Slow to put on and take off. |
| Users don't know exactly where to clip the microphone on. |
| Not good in a noisy environment. |
Desktop
Benefits:
| Moderate accuracy if it's positioned less than 30 cm from the mouth.
|
| User's hands are freed. |
| Cheap. |
Problems:
| The user has a "stick" pointing towards him, and usually
doesn't like this since it's visually distracting. Most users will reposition the
microphone so it points upwards, which reduces accuracy. |
| Users don't always know how to position the microphone. Some will put
it 90+ cm away, where accuracy will be poor. The best accuracy is when the microphone is
about 15 cm away. |
| Users have a tendency to lean into the microphone when they speak.
This is unnecessary. When users lean it adds extra effort to use speech recognition. Make
sure users know not to lean in. |
| Some users will lean in, and speak to the microphone with their mouth
about 3 cm away from the element. This produces poor accuracy. |
| Consumes desk space. |
| Not good in a noisy environment. |
Handheld
Benefits:
| Second only to close-talk and ear-piece microphones in accuracy
because the element is close to the mouth. |
| Some hand-held microphones are near-field element so they're noise
resistant. |
| Users have seen plenty of performers using these microphones and
already know how to use them. |
| Some handheld microphones have switches that allow the user to turn
speech recognition on/off. |
| Easy to pick up and put down. |
| User can lean back in his/her chair and talk. (Make sure the
microphone cord is long enough so users can lean back, even if their PC is under their
desk.) |
Problems:
| Users feel like a rock-star. |
| Users have to hold the microphone, which can get tiring if it's used
for long periods of time. |
| Microphone tends to roll when put down. |
Built into computer/monitor
Benefits:
| Nothing to plug in. |
| User's hands are free. |
| The user doesn't necessarily know where the microphone is so they
don't lean into it. |
Problems:
| Low accuracy unless the user is unusually close to the monitor.
|
| Not good in a noisy environment. |
Rest on top of monitor
Benefits:
| User's hands are free. |
| The user doesn't necessarily know where the microphone is so they
don't lean into it. |
Problems:
| Low accuracy unless the user is unusually close to the monitor.
|
| Not good in a noisy environment. |
Keyboard
Benefits:
| User's hands are free. |
Problems:
| Very low accuracy because of the typical position of the keyboard.
|
| Users might lean into the microphone to talk to it. |
| Microphone picks up keyboard clicks. |
| Not good in a noisy environment. |
In remote control
Benefits:
| Necessarily wireless. |
| If the remote is used to control an entertainment PC from the user's
couch, it's a good place to embed the microphone. |
| Remote can have a push-to-talk feature. |
Problems:
| Moderate accuracy if the user keeps the microphone about 30 cm from
his/her mouth. |
| Users have to hold the microphone, which can get tiring if it's used
for long periods of time. |
| Users don't know where to position the remote for best accuracy, so
they may lean into it. |
| Batteries. |
Additional Features
On/off switch, push-to-talk
Microphones can have an on/off switch built in that starts/stops
recognition. Alternatively, the microphone can remain off except when the user presses a
button down, talks, and lets go to stop talking.
Benefits:
| If the user has his/her hands already on the microphone, a button
provides a quick way to turn speech recognition on/off quickly and easily. Otherwise the
user has to hit a key on the keyboard. |
| Having a switch on the microphone is intuitive. |
Problems:
| Cost. |
USB
Most microphones plug into the PC with a 1/8" stereo mini-jack.
The A-to-D conversion is done internal to the PC. Alternatively, microphones can plug into
the USB port on newer machines.
Benefits:
| Users can't plug the microphone into the wrong place. Users sometimes
plug mini-jack microphone plugs into their speakers jack. The Microphone Setup Wizard
alleviates this problem. |
| Because the A-to-D is separated from the D-to-A, the PC can record
and play audio at the same time, even at different sampling rates. |
| Slightly better accuracy because the A-to-D is no longer in the
electrically noisy chassis of the PC. |
| USB microphones can have an on/off switch or push-to-talk that turns
speech recognition on/off. The on/off switch on analog microphones can only mute the audio
going to the speech recognizer, but the recognizer still has to process audio. |
Problems:
| Expensive. |
| Not all PCs have USB. |
Wireless
Any kind of microphone can be made wireless. Wireless microphone
have the following:
Benefits:
| No cords attached. |
Problems:
| Slightly lower accuracy
because of the RF transmission. |
| Batteries. |
| The user's
"conversation" with his computer is being broadcast, and is no longer private.
|
| Problems with several
wireless devices are used in the same area. |
| Expensive. |
|
