| Cognitive users access the cognitive radio networks in an opportunity way, however the random access of cognitive users and time-variant characteristic of channel quality make the available channel resources change quickly. Large bandwidth and low latency requirements of multimedia aggravate the variation of network. To improve the video quality at the receiver, we need to estimate channel quality dynamically. As a key technology of cognitive radio, spectrum prediction technology analyzes the sensed history information of the spectrum and predicts the availability of the channels to optimize the spectrum sense order, so that the sensing time overhead is reduced significantly. Otherwise, more important data is allocated on the more reliable channel.There are several well-known prediction algorithms based on Variable length Markov Model: PST?LZ78?PPM and CTW,which require the prefect sensing results. Otherwise, their prediction performance deteriorate significantly. Therefore, a Cooperative Spectrum Sensing Combined with Probabilistic Suffix Tree Algorithm(CSS-PST) is proposed to eliminate the influence of sensing error to prediction, which improves prediction accuracy through multiple user collaboration. Traditional hierarchical multimedia channel match strategies normally just consider the sensing-transmission time ratio but ignore the influence of some factors to the channel quality like noise and idle probability, which causes the lack of adaptability in complex cognitive wireless channel. We redefine the packet loss based on prediction results and channel noise to evaluate the channel quality. With the consideration of channel quality and the priority of layers, a Hierarchical-Matching(H-M) scheme is proposed to allocate the layered video data on multiple reliability-different CR channels. In this scheme, more important layer matches on higher reliability channel. Finally, the accuracy of CSS-PST algorithm and the effectiveness of H-M strategy will be proved through simulation. |
PDF Full Text Request