利用神经网络逆控制系统提高 Turbo码译码性能

对Turbo码译码逆模型建立问题,提出使用神经网络结构的非线性滤波器来建立Turbo码译码自适应逆模型.采用最优常系数比例因子统计得到Turbo码期望衰减系数,通过利用期望衰减系数训练神经网络非线性自回归外输入NARX滤波器,建立全局范围内的Turbo码译码逆输入输出映射模型.在线性逆控制系统中采用该自适应逆模型,与非线性逆控制结构的自适应逆控制系统相比,具有系统结构简单、运算量小等特点.仿真结果表明在信噪比大于0?dB时,该自适应逆模型算法收敛迅速、稳定,计算误差保持在较小的范围之内.自适应逆译码模型从译码机理角度提供了一种改善译码性能的新途径.

Using neural network inverse control system to improve the performance of turbo decoding

To solve inverse decoding model problem of turbo codes, an adaptive inverse turbo decoding model was proposed, based on neural network nonlinear filter.　Desired coefficients attenuation of cross-entropy (CE) was gotten by using optimal constant scaling factor with statistical principle. The auto-regressive exogenous input (NARX) neural network was trained to build the mapping model between the input and output of turbo inverse decoding all round by the desired coefficients attenuation. Compared with the nonlinear inverse control system nonlinear structure based,the proposed adaptive inverse model with linear structure has simpler structure, less computation. Simulations show that when the signal to noise ratio (SNR) is greater than 0?dB,the inverse model algorithm converges rapidly and stably with computational error being within a minimal bound. The proposed adaptive inverse model for turbo decoding has agreement with turbo decoding mechanism and provides a new way to improve the performance of turbo decoding.