Theoretical framework for evaluating partial checksum protection in wireless video streaming

The benefits of passing partially corrupted packets to the application instead of discarding them have been debated actively, since Lightweight User Data gram Protocol (UDP Lite) was introduced. UDP Lite allows partial check summing in order to omit bit errors in the non-critical part of the packet payload. Several studies have shown that data throughput over a link prone to bit errors can be significantly improved with partial check summing. However, the higher throughput comes at the cost of bit errors appearing in the non-critical parts of the payload. Therefore, the overall benefit depends highly on the capability of coping with errors at the application layer. In this paper, we present a theoretical framework for defining the optimal level of partial checksum protection, assuming that the bit error characteristics and the perceptual impact of bit errors appearing in the non-protected parts of the payload are known. We have also derived experimentally the distortion levels for video sequences coded with different bit rates and protection levels in the presence of bit errors. The results show that in some scenarios it is possible to improve the perceived overall video quality by using partial check summing.