改进的任意长度循环谱估计快速算法

平均的频域平滑算法( AFSM)是循环谱估计算法中数字频域平滑算法的改进，通过数据的加窗及重叠处理提高了循环谱估计性能。针对AFSM算法计算量大，对不同长度数据算法程序不一致的问题，利用算术傅里叶变换( AFT)计算DFT并利用一位谱相关( OBSCA)算法代替相关运算进行简化。给出了改进前后算法计算量的对比，并通过仿真验证了改进后算法的性能。理论分析和仿真结果表明，改进后的算法在估计效果基本不变的情况下，计算量减少，复杂度降低，且适合任意数据长度。

A Modified Algorithm of Arbitrary Length Cyclic Spectral Estimation

Averaged frequency smoothing method ( AFSM ) is one of the improved digital fre-quency smoothing methods of cyclic spectral estimation, which cloud increase the performance by window and overlap data. In order to lower the amount of calculation and the complexity, the arithmetic Fourier transform ( AFT) and the one-bit spectral-correlation algorithm ( OBSCA) are used to take the place of DFT and correlation. The calculation amounts before and after the improvements are deduced and compared. The performance of the improved method is proved by simulation. The theoretical analysis and simulation results prove that the calculation amounts of improved AFSM method reduce and the complexity of the improved method decrease with the performance almost unchanged. The proposed method is suit for data with arbitrary length.