ReDRESS

Non-acoustic speech recognition and reconstruction has a number of potential applications including speech rehabilitation and communication in noisy and covert environments. In this project we intend to focus on developing speech restoration systems appropriate for patients who have lost function in parts of the vocal apparatus, such as those who have had a laryngectomy as a result of throat cancer, trauma, destructive throat infections or damage to the nerves or those with other conditions which make speech difficult. Initially laryngectomy patients will be considered but in the long term, we would seek to consider a broader range of conditions such as dysarthria associated with cerebral palsy.

The essence of the proposed system, is that by monitoring the motion of elements of the vocal apparatus, it is possible to determine the speech sounds that an individual wishes to produce. An example of such a system might involve a number of small magnets implanted into the appropriate parts of the client’s mouth (eg lips, tongue and teeth). These will produce a variation in the magnetic field surrounding the mouth during ‘speech’. By monitoring these variations in the magnetic field and comparing them with training models of the signals for a chosen vocabulary, it is proposed that the best matching sound may be identified. It is envisaged that the implants would be small enough that they would not be apparent when observing the client and would not affect the movement of the mouth. The implants used are typically 1mm diameter adn 5mm long. These would be coated with a biocompatible material, eg silicone. Implants could be placed into the soft tissues by an injection technique under local anaesthetic. It is envisaged that the motion sensing system would be incorporated into a discrete device similar to a headset microphone or Bluetooth headset.

  Overview  Detail   Project Members   Publications

Copyright © 2011. Hull/Sheffield