本文给出了由频谱的相位函数重建原始信号的新方法。它能由相位函数得到各原始信号分量的振幅和延时值,以及频谱的模函数。新方法的叠代运算收敛快,因而其运算次数较以往的方法大大减少。本文还采用复指数展开算法,推广上述算法用于延时不是取样周期的严格整数倍的情况,进一步获得各原始信号分量的延时值的精确估计。最后,文中给出了计算机模拟实验的结果。 This paper presents a new method of reconstructing the original signal from the spectral phase function. It can be obtained by the phase function of the original signal component amplitude and delay values, as well as the spectral mode function. The iterative algorithm of the new method converges fast, so the number of operations is greatly reduced compared with the previous methods. In this paper, the complex exponential expansion algorithm is also used to promote the above algorithm for the case that the delay is not a strictly integer multiple of the sampling period, and the accurate estimation of the delay value of each original signal component is further obtained. Finally, the paper gives the results of computer simulation experiments.