Modeling And Simulation Of Turbo Decoding Structure Based On Linear Prediction

Turbo decoding structure is the main research object in this paper. First, the models for parallel turbo decoding structure, serial turbo decoding structure and turbo decoding structrure based on linear prediction are built using SIMULINK, and the performance is analyzed through the simulation. Compared with the traditional turbo decoding structrure, turbo decoding structrure based on linear prediction can reduce the decoding time and the algorithmic complexity through the linear prediction model. The linear prediction model is a linear polynomial model and can predict the extrinsic information in the iteration. Through the simulation results, we prove that the turbo decoding structrure based on linear prediction is available, and it can decrease the decoding time of the decoder after analyzing the running time of the linear prediction model and the decoding time of the decoder. After comparing the bit error rate performance of the turbo decoding structrure and the turbo decoding structrure based on linear prediction, we find that the former performance is 0.0~0.2dB larger than the latter one, and it is more close to the latter one while the iteration using the predictive result of the extrinsic information is larger. Then, the simulation results show that the iterative length is proportional to the decoding performance, the performance of Log-MAP algorithm is best in Log-MAP algorithm, MAX-Log-MAP algorithm and soft-output Viterbi algorithm.Then, we provide a sub-block parallel decoding structure based on linear prediction combining the linear prediction method and the sub-block decoding way, and analyze its performance through the simulation after building its model. The sub-block decoding structure of Turbo codes can lessen the decoding time dramatically while keeping good bit error rate performance through dividing the data into sub-blocks and decoding them simultaneously, but it needs more hardware resources. In this paper, we use several linear prediction models to predict the extrinsic information of sub-blocks respectively, so that we can reduce the decoding time of the sub-blocks and lessen the decoding time of turbo decoder. Then we build the sub-block parallel decoding model of Turbo codes and the sub-block parallel decoding model based on linear prediction. Through the simulation, we prove that the sub-block parallel decoding model based on linear prediction can reduce the decoding time, compared with the sub-block parallel decoding model of Turbo codes. The simulation results show that the performance of the sub-block parallel decoding model based on linear prediction is proportional to the sub-block length, and inversely proportional to the parallel degree. And it needs more hardware resources because of the linear prediction models.