Bristol Wearable Computing

[ Home | Plans | Meetings | Members | Search | Papers | Links | CyberWear | LocoSoft]

Speech Engine - Application Grammar v 2.4

To meet the end-Feb target for a demonstration of the speech engine with different users in different environments, a simple interim grammar is proposed for the speech interface. It consists of a log in username; seven main commands; and nine sub commands/simple responses to queries from the computer.

In this document words in bold and double quotes signify user speech commands e.g. "start"; and words in bold and italics signify responses/queries from the  computer e.g. done.

Start Sequence (Logging On)

The operating system opens with the computer saying Welcome to the CyberJacket! The computer then asks Say your name to start the user responds with a unique "username". If recognised the computer requests the user to say "start".

If the "start" command is correctly recognised the computer responds: Jacket's away! Which application would you like?

If either the "username" or "start" is NOT recognised the computer will say Sorry can't identify you, known users are and list the known users.

n.b. The use of "start" ensures that the jacket hasn't mis-recognised the "username". Otherwise it could be  possible to log on as the wrong person and then be unable to exit as the users have different voice profiles.

Command Grammar

Having successfully logged on the user has a choice of applications and utility functions which may be run. The most complex of these is the NoteBook which is described in a separate section below. The others are:-

  1. "time" - The computer responds with the current time: The time is <hh mm>
  2. "position" - The co-ordinates of the user are played: you are at: latitude ## degrees ## mins ## secs N/S, longitude   ## degrees ## mins ## secs E/W .
  3. "space" - The remaining space for notes is quoted by the computer.
  4. "say again" - The computer repeats it's last message.
  5. "shutdown" - The computer requires the user to confirm with "yes", "no" or "exit". If "yes" the linux shutdown procedure is initiated and the user is informed of it's completion with the message: It is now safe to turn off your computer. Otherwise the response is: Done.
  6. "help" - All user commands in the current menu are replayed by the computer.
  7. "exit" - Logs off the current user, the computer acknowledges with: Jacket sleeping.

The NoteBook Application

This application enables the user to 'open' the pages of a notebook (or domain) and 'read' the notes on a particular page. Once the notes have been played, the user has the option to add further notes. For the purposes of the 'February demo' the notebook has been restricted to three pages - "shopping list"; "to do list"; and "tourist notes". In the longer term it will be possible to make the pages context sensitive - only (dis)playing the notes which are relevant to a particular location as originally conceived in LocoMedia, or potentially to aid a user who is experiencing a particular emotion. 

"notebook" (Visit domain)

  1. If the user is already in the notebook (visiting a domain), a message will say 'You are currently on page  <xxx> Would you like to go to another page?'
    1. If the user says 'no' or 'cancel' the jacket tells the user that the operation has been cancelled, and this function is aborted.
  2. ??Pre-recorded message?? ask the user to specify a page name.
  3. Jacket recognises the page.
  4. A call to the domain server temporarily clears the domain list, and then adds the named domain to the subscribed list.
  5. Notes recorded on the page are automatically played.
  6. The user can check how much space is left in the NoteBook, record another note or close the notebook:
    1. "space" - Allows users to find out how much storage space remains in their NoteBook. The computer replies: There is space for another <n> minutes of notes or when less than a minute remains: There is only space for another <n> seconds of notes.
    2. "record" -  The computer responds: Recording note; the 'Jacket beeps to prompt user to record free audio; and a metronome style click plays while recording the note until the user says "end of note" or "cancel". If the user said end-of-note, the note is compressed, stored and added to the NoteServer. Attributes for the current users contexts are added to the notes (sensed and user).The jacket uses a pre-recorded audio message to say that the note was stored without problem, or that the note was discarded depending on what happened. If the user said cancel the computer responds: Recording cancelled and the recording is aborted.
    3. "close" (?? Pre-recorded message ask the user to specify a domain name. The 'Jacket recognises the said domain.A call to the domain server restores the users subscribed domains to what they were before the visit started ??). The computer responds: Done - returning to start menu. Which application would like?

Other Computer Responses

  1. Sorry, I don't know! In response to an utterance that it could not understand or that didn't comply with the grammar described above.
  2. Sorry, there is no more space for new notes! When a note has to be cut short or rejected entirely through lack of space.
  3. Sorry, that was not understood | I didn't recognise that | Sorry, not recognised Said when an input is not confidently recognised.

The above syntax is illustrated in the diagram below:-

The menu boxes are labelled in green; user commands are shown in the menu boxes in italics; and the Cyberjacket's actions are shown in blue.

Stop Press - Developments!

1) Two new commands have been added - "Sleep" and "Wake up". The CyberJacket snores briefly - zzzzzz - after being told to sleep, and says Awake after being told to wake up. While it is asleep it will not respond to any other commands, though it may randomly snore (in response to snore-like commands).

2) Using the 'Jacket in different environments has introduced the need for the dynamic control of record/playback levels. The commands "Louder" and "Softer" have been implemented to vary the earpiece volume in approximately 3db steps between minimum and maximum levels. Work continues on the best method to control record levels (manual/automatic?).

unicrest.gif (4191 bytes)

The material displayed is provided 'as is' and is subject to use restrictions.
For problems or questions regarding this web contact Cliff Randell.
Last updated: January 14, 2000.
logoep.gif (1404 bytes)
ęCopyright Hewlett-Packard 1997-2000.