P2P Video Multicast for Wireless Mobile Clients Johannes Karlsson and Haibo Li Digital Media Laboratory (DML) Dept. of Applied Physics and Electronics Ume˚ a University, Ume˚ a SE-90187 Sweden {johannes.karlsson, haibo.li}@tfe.umu.se Jerry Eriksson Department of Computing Science Ume˚ a University, Ume˚ a SE-90187 Sweden jerry@cs.umu.se Abstract— Mobile TV is a new interesting area in the telecom- munication industry. The technology for sending live video to mobile clients is characterized by relatively low CPU processing power, low network resources and low display resolution. In this paper we discuss a solution to all of these problems by using application layer multicasting. This can significantly reduce the needed bitrate and required computing resources for each client. At the same time the received video quality is increased. Several different methods for splitting the video into substreams are discussed. A system for application layer multicasting using layered H.264 is also presented. I. I NTRODUCTION High-bandwidth live media distribution is very important in many applications, ranging from mobile entertainment to mobile services of industries where for example a technician needs to access 3D visualization of dynamic processes. The technical challenge lies in the fact that mobile clients have very limited network resources, about 30 kb/s for GPRS and 100 kb/s for 3G networks, which greatly limits their ability to access the high-quality media information. The solution is to let the team make use of short-distant wireless communication to build a cooperative environment and multicast the live content within the team. Unfortunately, traditional IP multicast system doesn’t work. An alternative is to use the application- level multicast. The idea is that the mobile hosts pool their resource to route and distribute multicast messages using only unicast network service. The local wireless communication can be built up using for example Bluetooth or IEEE 802.11. To extend the coverage a multihopping ad-hoc network can be used. Here all nodes also act as routers and can relay information for each other. In this network the nodes are free to move randomly and organize themself arbitrarily. This is a fully autonomous system that works without preexisting infrastructure. Ad-hoc networks has been an active research area for a while and several protocols has been proposed. Two examples of the most commonly used ad-hoc routing protocols are AODV [1] and DSR [2]. A Peer to peer application layer multicast approach for sharing files on Internet has been used for some time. One of the most commonly used protocol for this is BitTorrent [3]. Peer to peer media multicasting has also been considered in several papers [4] [5] [6]. None of these do however consider the case for video distribution for mobile handheld devices. These devices are characterized by relatively low CPU processing power, low display resolution and low bitrate. Fig. 1. Overview of the system The video must be split into several substreams to make it possible to use application level multicasting. This can be done in several different ways. The two most commonly used meth- ods are multiple description coding and layered coding [7]. The multiple description coding is very flexible. The quality of the received video improves with each received description. The total bitrate needed for the multiple description system to achieve a given distortion will however be relatively much higher than the bitrate needed to achieve the same distortion using a single stream. Layered coding can achieve both spatio- temporal and quality scalability. The coding efficiency for this technique is also relatively high. The drawback of layered coding is that the decoding process requires rather much computing resources compared to single layer coding. This is a major problem for the mobile clients that are usually battery powered and have very limited computing resources. Another method for creating substreams of the video is to divide the video into areas where the areas are coded to independent streams. The first user in a group will receive the most important area. If there are two users the two most important areas should be transmitted, and so on. This will require very low computing resources since each user only decodes a part of the entire video. II. SCALABLE H.264 FOR P2P MULTICAST Assume that a group of users are watching the same real- time video over a GPRS connection. We also assume that these users are equipped with a local wireless connection, for example Bluetooth or 802.11. See figure 1 for an overview of