DEVELOPMENT OF A PAEDIATRIC AUDIO-VISUAL SPEECH TEST IN NOISE

Arnold L (1), Canning D (2), Dykmans P (1), Vanpoucke F (1).

(1) Advanced Bionics Europe; (2) Hear2Learn Ltd.

Background: The importance of audio-visual speech integration in the communication of children with cochlear implants may sometimes be underestimated. Not many tools are available to assess lip-reading abilities.

Aims: To develop a paediatric audio-visual speech test to evaluate the benefit obtained from adding lip-reading information to auditory signal.

Methods: Materials from the existing McCormick and English as a Second Language tests were selected. The words available were recorded from male and female children. Editing was performed to add competing noise to the speech signal. A system capable of driving two screens simultaneously (“subject” screen allowing lip-reading and “clinician” control screen) was created and software prepared to drive it. Pilot testing was performed on twelve paediatric cochlear implant users, first with lip-reading, then with auditory signal only.

Results: The video recordings provided acceptable quality. The preliminary version (PAVT 1.0) of the software platform was tested: words were randomly presented to the subjects whose task was to indicate the corresponding picture or toy, or to repeat the word. Testing may be conducted in quiet or noise, with adaptive signal to noise ratio. The clinician may select audio only or lip-reading with audio. Controls were available for scoring and automatic report generation. The results showed differences in scores obtained in the audio-visual condition compared to the auditory only condition, highlighting the benefits that might be obtained from adding lip-reading information to auditory signal. Feedback was collected on how to improve the interface and a new version was released (PAVT 2.0), which will be introduced, together with some supporting results.

Conclusions: The feasibility of using recorded audio-visual material to assess lip-reading abilities was confirmed. The option to test in noise allows for a better representation of real-life conditions. The next steps will involve validation of PAVT 2.0 on a larger group of subjects.

E-mail: laurea@abionics.fr