Research On Efficiently Iterative Decoding Algorithms Based On Stochastic Computation

The goal of the future mobile communication system is to meet the demands for massive connections,huge amounts of data and green communications.To achieve this purpose,the non-orthogonal multiple access technology is considered a competitive alternative to replace the orthogonal frequency division multiple access technology to improve the spectrum efficiency and the ability of connection.Sparse code multiplex access(SCMA)technology is one of non-orthogonal technologies which is believed the multiple access technology for the next generation mobile communication system.In the prospective mobile communication system,higher-order modulation will be a method to realize high-speed communications.However,the computation complexity of the best performance detection algorithm of the SCMA detector is increasing exponentially with the modulation order,which will be a serious problem for SCMA technology to eventually become a physical standard.Especially,in order to realize green communication,the implementation of the communication system under the scenarios of massive connection and low power consumption will be a huge challenge.To cope with the demands for massive connections and low power consumption in the future wireless communication systems,in this thesis,a variety of low complexity and low power consumption detection and decoding algorithms are proposed for Turbo coded SCMA systems,The contribution of this paper are summarized as follows.1.A fast convergence and low hardware consumption stochastic Turbo decoding algorithm is proposed.Firstly,to solve the slow convergence caused by the scaling stochastic addition in traditional stochastic Turbo decoding algorithm,this thesis proposes a high accuracy no-scaling stochastic addition to improve the convergence.Secondly,to reduce the hardware consumption and improve the accuracy of the stochastic normalization unit,a novel stochastic normalization unit is proposed.As a result,the novel stochastic Turbo decoding algorithm with the high accuracy no-scaling stochastic addition and novel stochastic normalization unit,improves the convergence speed by about 7 times and reduces the hardware consumption to 55% of traditional exponential stochastic Turbo decoding algorithm.2.Three low computation complexity detection algorithms are proposed to reduce the computation complexity of the traditional SCMA detection algorithm.Firstly,a detection algorithm named JUP-MCMC algorithm based on Gibbs sampling is proposed,which overcomes the drawback of the traditional SCMA detection algorithm computational complexity increasing exponential with the modulation order.The computational complexity of the proposed JUP-MCMC algorithm is only 4% of the traditional SCMA detection algorithm when the modulation order is 64,and the hardware consumption can be further reduced by about 35% when multiple partition decomposition stochastic multiplier is adopted in JUP-MCMC method.Secondly,two kinds of stochastic SCMA detection algorithms including binary-domain stochastic SCMA detection algorithm and multivariate-domain stochastic SCMA detection algorithm are proposed.The hardware consumption of proposed stochastic SCMA detection algorithms is 9.88% and 10.85% of the traditional SCMA detection algorithm with no performance loss,respectively.3.The joint detection and decoding algorithm is proposed and its implementation scheme based on stochastic computation is also proposed.The traditional iterative detection and decoding scheme has many drawbacks,for example,the low convergence speed,the huge computational complexity and the performance loss compared with the orthogonal scheme.The proposed joint scheme is based on the joint factor graph,and it can approximate the system performance of the orthogonal multiple access scheme when the computational complexity is appropriately increased.Moreover,as the joint scheme has faster convergence speed,thus,the computational complexity is only 30% of that of the traditional iterative detaection and decoding algorithm with no performance loss.Meanwhile,the joint scheme has 1.2d B performance gain when the computational complexity is increased to 60% of that of the iterative scheme.To further reduce the hardware area and power consumption,the joint scheme is implemented based on stochastic computation,where all arithmetic operations are implemented in logic gates.