Video Signal Of The Network In Real Time Chaotic Secure Communication

Chaotic secure communication is recently the hot research topic of cryptology.It consists of many facets to which network chaotic secure communication is one of the particularly promising research directions. This dissertation realizes the real-time video' s network chaotic secure communication system based on the chaotic cryptology theory, network communication theory and H263 encoder&decoder. The system produces real-time video secure communication between service and client. In the meantime, we got several achievements during the process:Firstly, a sequence encryption algorithm of real-time video in use of one dimension piece-wise linear chaotic map is carried out. This algorithm makes use of Inter' s SSE2 special dictates and makes nonlinear change of chaotic sequences.As a result, the algorithm not only encrypts faster, but also resists several cipher analysis methods. Secondly, this paper presents a block encryption algorithm of real-time video based on chaotic neural network which makes use of numerical transient chaotic neural network to produce chaotic sequences and uses these sequences to do encryption. The algorithm has vast amount of ciphers and higher security. Furthermore, because of computing in parallel, it can realize in hardware easily. Thirdly, we sends the numbered cryptographic H263 video frames based on the RTP/RTCP protocals. At service endpoint, based on the received SR and RR packages, we either change the capture rates or change the rates of I frames to P frames to ensure the quality of video sevices. Last but not the least,this paper presents a hyperchaotic prediction algorithm using improved NRBF network. With the NRBF network architecture based on resource allocation thought, the algorithm deletes or adds hidden-layer neurons dynamicly, and what is more, it uses Levenberg-Marquardt algorithm to train the centres and widths of the neuron and the weights of output layer simultaneously. At last, we do off-line prediction with this algorithm and the results show that it can predict well in a short period.