IMPROVED SPEECH RECOGNITION THROUGH VIDEOTELEPHONY - EXPERIMENTS WITH THE HARD-OF-HEARING

Three experiments were carried out to assess the potential effectiveness of videotelephony as an adjunctive aid to speech reception by the hard of hearing. The experimental procedure consisted of audio-only and audio-visual presentation of standard prerecorded sentences to hard-of-hearing subjects who subsequently had to repeat as many words as possible. The percentage of correctly reported syllables was taken as the speech reception score. Several independent variables related to picture quality parameters. The most important of these were the temporal resolution (frame rate) and the spatial resolution of the video image. The spatial resolutions were QCIF (180 x 144 pixels) and 1CIF (360 x 288 pixels), and for both these resolutions the image was processed through a 64 kb/s codec. The results showed that the speech readability of a video image improves as the frame rate is increased to 15 Hz and that a further increase in frame rate does not result in a further improvement of speech readability. Audio-visual presentation of 64 kb/s coded images with either QCIF or 1CIF led to significantly better speech reception than audio-only, but the scores were lower than the scores for comparable analog broadband images. Implications of these findings are: 1) 64 kb/s videotelephony is potentially effective in improving speech reception for the hard of hearing; 2) In the design and selection of video codecs for this purpose, a frame rate of 15 Hz is recommended; 3) Both QCIF and 1CIF should be regarded as suitable resolutions.