Auditory-Visual Speech Processing (AVSP'98)
December 4-6, 1998
An experiment was conducted to investigate the potential benefits of a visual-FM system for speechreading. Two speakers took part in the study. They were recorded while they spoke 36 sentences at a distance of 1.83, 3.66, and 7.32 m from a stationary Hi8 video camera. Under each experimental condition, the signal available from the camera of a visual-FM system was recorded simultaneously. A speechreading test consisting of a randomization of 432 recorded sentences was administered to a group of 16 subjects with normal hearing and normal (or corrected normal) visual acuity. The results revealed that for the recordings obtained with the Hi8 mm camera speechreading performance decreased as a function of distance. Specifically, there were no differences between the recordings made at 1.83 and 3.66 m. However, there was a significant difference between those two recordings distances and the recordings made at 7.32 m. For the recordings obtained with the visual-FM camera, speechreading performance did not vary significantly as a function of distance. Those findings indicate that the visual-FM system constitutes an effective method of providing visual-speech cues in environmental conditions where the distance between the speaker and the speechreader is not optimal for live-speechreading.
Bibliographic reference. Gagné, Jean-Pierre / Monday, Kim Le / Desbiens, Christine / Lapalme, Marie / Ducas, Luc (1998): "Evaluation of a visual-FM system to enhance speechreading", In AVSP-1998, 167-170.
|av98_167_1.mov (24631 KB)||visualfm.zip||Demonstration video of Visual-FM system to enhance speechreading||Video File: QuickTime; Video Track: Size 240x180 pixels, Duration 04:03:06, millions colors, Data Format: Sorenson Video, - Sound Track: Sample Rate: 22.05kHz, Mono, Sample Size: 16 bits, Compression IMA 4:1|