Prototype Testing - Voice Recognition Scenarios

For this test, we decided to see how personas would use the microphone to input commands, and compare results. We asked them to say how they would instruct the device to perform each scenario.

Maureen

Logging Weight Information - "Add my weight... ummm.. its 12 stone seven."

Viewing Weight Information - "View my weights? This is strange..." (she forgot to let go of the microphone button).

Viewing an Unread Message - "View my messages... ummmm..." This task couldn't be pweformed in one go as the user needs to know what the unread messages are.

Modifying Date and Time - "Change the date to the 18th of March and the time to 2:50" A problem here is that the time could be 02:50 or 14:50.

Turning on Security - "Turn my security on please." She felt a bit silly saying "please" afterwards.

Maureen eventually got the hang of the speech input, but it seemed a little odd at first, and possibly made her feel a bit embarrassed.

Fiona

Logging Weight Information - "Insert my weight, 49 kilos."

Viewing Weight Information - "Show my weight."

Viewing messages - "Messages."

Modifying Date and Time - "Date, 18th of March, 2 fifty."

Turning On Security - "Security on."

Fiona was very comfortable with the speech input, and she liked the efficiency of the process. She also perceived it to be a very high-tech feature, and something she could show off. One important point to note is the use of kilos, the device must be able to handle different units.

Derek

Adding Weight Information - "Hello? Do I speak into here? ADD MY WEIGHT... TWELVE STONES AND SIX."

Viewing Weight Information - "SHOW ME MY... I don't need to shout? Why can't I do this the other way?"

Viewing Messages - "Show me the messages. Is that okay?"

Modifying Date and Time - "18th of March, five and twenty past three."

Turning on Security - "Turn on security. Is that the last one? Thank goodness for that. Oh, I haven't let go..."

Derek found this method very odd, even more odd than using the touch screen. One thing to notice is the old-fashioned way of saying "twenty five past three".

Conclusion

The speech input mechanism was not effective for all personas, but it did achieve its goal of efficiently performing tasks in some cases. For it to be effective, the speech recognition software needs to be able to filter out junk words and be able to handle the numerous ways of saying the same thing. Another important point is that the user needs to know when they can use it, and clear notification needs to be provided if more information is required to perform a task.