Visual perception of content-prioritised sign language video quality.

Video communication systems currently provide poor quality and performance for deaf people using sign language, particularly at low bit rates. Our previous work, involving eye movement tracking experiments and analysis of visual attention mechanisms for sign language, demonstrated a consistent characteristic response which could be exploited to enable optimisation of video coding systems performance by prioritising content for deaf users. This paper describes an experiment designed to test the perceived quality of selectively prioritised video for sign language communication. A series of selectively degraded video clips was shown to individual deaf viewers. Participants subjectively rated the quality of the modified video on a Degradation Category Rating (DCR) scale adapted for sign language users. The results demonstrate the potential to develop content-prioritised coding schemes, based on viewing behaviour, which can reduce bandwidth requirements and provide best quality for the needs of the user. We propose selective quantisation to reduce compression in visually important regions of video images, which require spatial detail for small slow motion detection, and increased compression of regions regarded in peripheral vision where large rapid movements occur in sign language communication.