Progress with action points from STARDUST meeting 27th April 2001

Joint meeting with stardust participants and external reviewer.

  1. Mark Tataham’s role defined to include:

  1. External review of progress and ensure the project stays within the boundaries for which ethical consent has been granted.
  2. Expert advice and guidance.
  3. Produce a report indicating the current progress for the DoH (to be agreed between Mark T and DoH).

It is suggested that external review meetings take place in Sheffield approximately every 3 - 4 months.

Agenda point

  1. Tony Horn from DoH to be contacted regarding the possible payment of Mark Tatham for producing the review reports. (Expenses to be meet from available resources). Mark H and Mark T
  2. Agenda point

  3. DoH research governance details to be forward to Mark Tatham – MRC, DoH -
  4. Completed

  5. Speech consistency to be measured automatically by acoustic vectors (per frame). However, this will not analyse the duration of the utterance and this must be considered in the definition of consistency. This may be combined with measures based on speech therapists perceptions. KTH should be contacted to see if they have done similar work. Phil and Nassos
  6. Agenda point

  7. Status of recruitment of participants: 5 accepted and initial data gathered, 1 in the process of joining, 2 possible. Recruitment to continue until 12 participants have been recruited. Mark P
  8. Agenda point

  9. Data gathering to continue, including ‘fast track’ data of the ECS and communication words, plus the 10*10 words for the consistency score for 2-3 people. Mark P
  10. Agenda point

  11. Appropriate methodology for the picture description data to be defined. When the methodology has been accepted, intelligibility scores should also be obtained. Mark P and Pam
  12. Agenda point

  13. When the prototype trainer is available provide support and assistance to users. Feedback from users to be noted for future reference and software developments. Mark P
  14. Outstanding

  15. The vocabulary for the training software is to be the ECS and communication words defined appropriate for each individual user by Mark P. (Therefore, in the order of 10-20 words). Notes for software: Nassos

    1. the system is to be based on two levels, i.e. level 1) ‘television’, ‘fan’ etc. Level 2) ‘switch on’, ‘switch off’.
    2. During training, the user needs to be prompted to breathe between levels, therefore ensuring that the 2 utterances are in isolation. The pause also acts to ensure the most consistent speech. I.e. ‘Television’, pause, ‘switch on’.
    3. Improvements in speech are to be measured by the consistency software produced in point 4. (All spoken input during training to be recorded and kept for future reference)
    4. Feedback to be provided to the user both visually and audibly, so that they can compare their utterance with the target.
    5. The target utterance for each user is to be their best attempt at a word (obtained prior to training). Therefore, whenever they use the training software it should report a score to the user that compares the consistency of the spoken utterance with the target utterance. If it is evident that speech is improving then it may be necessary to change the target utterance to the ‘best’ utterance obtained during the training phase. If this occurs then the software may be improving speech consistency and intelligibility.
    6. Ultimately it is envisaged that the trainer should operate in windows and function using a switch rather than a mouse. Note BDGH has software (works in Dos and windows 16 and 32 bit) that continuously poles the serial port (pins 4 and 8) and monitors when a switch is pressed and released. This code can be made available to the project.

  1. Appropriate speech recognition tool kits should be discovered and a review of their strengths and weaknesses conducted. Based on this review a recommendation can be made and discussed at the next meeting. Nassos and Phil
  2. Agenda point

  3. Organise review group meeting in last week of May (Mon/Tue for Mark P). At Pams office. Simon
  4. Completed

  5. Organise review group meeting with the external in last week of June (Mon/Tue for Mark P). At ILSAH. Simon
  6. Completed

  7. Forward relevant papers and information to Mark Tatham. Simon
  8. Completed

  9. Purchase additional laptop computer. Nassos and Simon
  10. Completed

  11. Facilitate the repair of Mark P’s laptop. Nassos
  12. Completed

  13. Purchase another array microphone, so that both laptops can be left with clients if necessary. Simon

Awaiting delivery

Major deliverables over the next 2 months (to be completed by July 1st)

  1. Modify OLT to work with the 10-20 words for training. It should provide both visual and audio feedback. During this phase it is envisaged that the program will work on the linux platform and be transferred at a later date to windows. The resulting software should be tested with users and feedback obtained by the end of May.
  2. Evaluation of speech recognition tool kits to be performed and recommendations made for the next review meeting (early June). The software can then be purchased and work can commence late June.
  3. Software for obtaining a consistency score should be created as a stand-alone package and incorporated into the trainer by the end of June. (At some stage it may be appropriate to combine the speech recorder and consistency software as a tool for speech therapists, freely downloadable from the web-site?)
  4. A software interface should be written to enable data obtained from the recorder to be easily played back. Anticipated during June.

Other work

Paper presented at Workshop on Inovation in speech processing. WISP 2001. Victoria Hotel. Straford-upon-avon. England 2nd-3rd April. "A two dimensional Kinematic mapping between speech acoustics and vocal track configurations." – Hatzis A. Green PD.