Efficient layered coding for heterogeneous networks

The problem

In today's scenario diverse data consumption devices (e.g., smartphones, tablets, smart-TVs, PC) all access the same multimedia content (e.g., live event broadcast) over networks of time varying bandwidth and latency. This implies, the same content is served at different data rates and quality levels to various users. Obviously all the different quality levels of a signal share some common information between them. Thus it is critical to optimally exploit this shared information to ensure efficient use of resources for storage and transmission across the network. The simplest solution would be storing and transmitting independent copies of the signal at different quality levels, but this is highly wasteful in resources. In an alternative approach, all the information conveyed in a low quality version is forced to be part of the high quality version. This rigid hierarchical structure, popularly known as scalable coding, is widely recognized to be significantly worse in quality compared to individually encoded copies.

Our solution

Our approach is to enable routing to a high-end receiver only a subset of the bits that are sent to the base receiver. This relaxation of the hierarchical structure provides enough freedom for a well optimized design to achieve best attainable quality at both the receivers, yet significantly reduce the transmission and storage rates, and also provides the opportunity to control the layered coding penalty and achieve intermediate operating points in terms of the trade-off between total transmit rate and total receive rate. We have developed solutions based on this relaxed hierarchical framework by first developing the building block of quantizers and then using them to completely overhaul the layered coding approach for audio and video content.