Exploiting multiple description coding for intermediate recovery in wireless mesh networks

  • Authors:
  • Pradipta De;Nilanjan Banerjee;Swades De

  • Affiliations:
  • IBM Research, India Research Lab, New Delhi, India;IBM Research, India Research Lab, New Delhi, India;Electrical Engineering Department, Indian Institute of Technology Delhi, New Delhi, India

  • Venue:
  • Journal of Network and Computer Applications
  • Year:
  • 2014

Quantified Score

Hi-index 0.00

Visualization

Abstract

Guarding against channel errors in wireless networks has been a challenging research problem, specially when transmitting time-constrained contents, like streaming video or image. Multiple description coding (MDC) is an effective means to combat bursty packet losses over wireless channels, and it is especially promising for the applications where retransmission is unacceptable. In MDC, a frame is broken into several equally important descriptions which can be sent over multiple paths to the destination. Meshed multipath routing over wireless mesh networks and MDC, when combined together, could make video/image transmission more error-resilient by providing opportunities to recover erroneous video packets at the intermediate nodes along the route. In this work, we propose to combine MDC and meshed multipath routing in such a way that multiple descriptions are sent over different paths in the network and are merged at the intermediate nodes for possible recovery of corrupted descriptions. The routing uses the idea that when multiple descriptions join at intermediate points it can help in partial recovery of lost or corrupted descriptions by using the link error information. This approach reduces the possibility of error propagation, thereby improving the overall image/video quality at the destination. In order to achieve high gain in terms of peak signal-to-noise ratio (PSNR), we investigate an optimum spatial interleaving based MDC scheme that maximizes the intermediate recovery possibility for a given recovery filter design. We also explore the choice of an optimum number of descriptions for merging at an intermediate merging point that maximizes the PSNR gain. Our simulation study shows that, by introducing recovery possibility even at one intermediate node, it is possible to achieve a PSNR gain of around 5-6dB using our coding, routing, and recovery strategies.