基于共轭梯度求解代价函数的卷积码参数识别算法

卷积码常作为级联码、Turbo码等高性能编码的子码，正确识别出卷积码的参数是级联码、Turbo码参数识别的基础，这要求卷积码参数识别算法具有较强的抗噪能力。利用解调软判决序列可以有效提高识别算法的抗噪能力。根据递归系统卷积码编码码元间的线性约束关系构造了一个基于指数函数的代价函数模型，将生成矩阵的识别问题转化成求解代价函数极小值的最优化问题，并采用共轭梯度法不断逼近极小点。仿真结果显示，与现有算法相比，所提方法显著提高了抗噪能力，且适用性强、收敛速度快。

Parameter estimation algorithm of convolutional codes with solving cost function based on conjugate gradient

Convolutional codes are often used as sub-codes of high performance codes such as concatenated codes and turbo codes. Correct parameter estimation of convolutional codes is the basis of recognition of concatenated codes and turbo codes, which requires that the estimation algorithms of convolutional codes should have strong robustness against channel noise. The key to such purpose is to make use of the soft-decesion demodulation received sequence. In this paper, a cost function model based on exponential function is proposed according to the linear constraint relation between symbols of recursive systematic convolutional codes. The parameter estimation of convolutional codes is transformed into the minimal value of the cost function. And the optimization is accomplished via a simple iterative process by conjugate gradient. Simulation results show that, compared with the existing algorithms, the new algorithm significantly improves the performance while it is also applicable to the estimation of general convolutional codes and has a fast convergence speed.