Real-time Multimedia Transmission Over Heterogeneous Networks

The development of mobile Internet makes it become popular around the world. However, the functionalities of it can still not meet the requirements of people, especially the realtime communication of multimedia. Accordingly, how to ensure realtime multimedia communication over mobile Internet becomes an academic hot spot. Meanwhile, with the existing of other hetergeneous networks, such as Wi-Fi networks, WiMax networks, PSTN networks, GSM networks and 3G networks, the multimedia communication among these hetergeneous networks becomes increasingly important. Although there exists many optimal algorithms aiming at improving performance of multimedia transmission, the communication bewteen diferent networks is still a bottleneck, specially the performance and quality of packet transmission on the IP networks.Based on the above problems,the works of this thesis mainly focus on optimization and control of multimedia transmission, as well as the performance and quality. We firstly analyze the current works on multimedia networks communication and the standards to evaluate performance of transmission. And then, we propose a Port Correlation Analysis and Prediction (PCAP) algorithm to complete NAT traversal using TCP. While transmitting multimedia between different networks, we propose Mean Average Prediction Strategy (MAPS) and Gradiant Prediction Strategy (GPS) to make the handoff bewteen heterogeneous networks smoothly and seemlessly. After that, the technology of silence detection is employed. We devide the video frame into several small pieces and encapsule these pieces into the silence voice frames. These encapsuled voice packets are then transmitted to the receiver side. In order to make efficient use of bandwidth to improve quality of media playout and reduce the network conjection, we adopt Binomial Probability Model (BPM) to control redundant packet number and sending rate. Finally, we implement a multi-point conference prototype system. In the system, we use adaptive probability mixed-voice scheme and conference management to enhance the efficiency of mulitpoint conference system. All of these research works will serve as an important contribution in the field of multimedia transmission among heterogenerous networks. More details of our contribution can be described as follows:(1) We first compare and analyze those approaches used to solve NAT traversal problems in P2P systems. For TCP NAT traversal, we propose a Port Correlation Analysis and Prediction (PCAP) algorithm to predict the NAT allocated port number before establishing a TCP connection. A novel PCAP based TCP NAT traversal scheme is then proposed for building a successful P2P connection. We test our method through experiments under real networking environments. The results demonstrate the effectiveness of PCAP algorithm on NAT behavior detection and port prediction. Also, using PCAP, the system can achieve high successful P2P connection rate.(2) Aiming at avoiding the drawbacks of IEEE802.11 standards on wireless handoff, a Moving Average strategy based on Neighbor Graph (NG) is proposed. Through employing background scanning to collect signal strength of current Access Point (AP), the strategy dynamically calculates the mean value of signal strength. It then decides to handoff to the candidate AP according to the mean value. In order to decrease the loss of data packets, the strategy forces the packets to be transmitted to candidate AP by querying the NG of current AP. Besides, we analyze the characteristics of handoff from 3G networks to WLAN and propose Gradient Prediction Scheme (GPS). GPS starts the procedure of handoff through computing the first derivative value of the received signal strength. The combination of MAFS and GPS improve the efficiency of vertical handoff and abbreviate the disconnecting time.(3) Using silence detection technology, we propose silence-based adaptive voice and video transmission strategy. This strategy divides the video frame into several small pieces and packed these pieces into silence frames. The packed packets are sent to the other peer. Meanwhile, One way delay algorithm is employed in order to evalute the available bandwidth of network links timely. We use OWD result to adaptively adjust the sending rate and reduce network conjection. For the purpose of improving quality of voice and video transmission, we analyze the related works of loss recovery technology and propose Binomial Probability Model(BPM), which is based on Forward Error Correction(FEC) Reed-Solomon codec technology and interweaving loss recovery technology. BPM calculates the number of redundant packets according to the feedback from the receiver. By using interweaving technique to transmit mixed-packet of voice and redundancy, BPM can save bandwidth effectively.(4) An efficient design and implementation of control and multiplexing protocols H.245 and H.223 is proposed, which is an important part of 3G324M protocol stack, for mobile wireless video conferencing. Our implementations aim to support the multipoint video conferencing with the capability of transmitting/receiving multiple video/audio streams simultaneously. To enhance performance of our system, we also propose a novel and simple fast real-time adaptive cross-platform multi-party conference scheme. In the proposed scheme, priority is decided by probability. That is, the server calculates its audio probability according to energy value and the coded data size of each client, then decides which is the current spokesperson by the audio probability, finally transmits voice and video packets.