综合采用直接序列扩频(DSSS)和跳频扩频(FHSS)体制的直序/跳频(DS/FH)通信信号显著降低了单位时频面积上的功率，经典的时频-多信号分类(TF-MUSIC)算法对这一类信号进行入射角(DOA)估计时误差较大。为解决该问题，本文对TF-MUSIC 算法进行改进，在时频二维平面上计算DS/FH 信号的时频特征值，并设计相应的时频掩膜滤波核函数，再根据估计出来的循环平稳频率计算相应的循环平稳空间自相关矩阵，得到了循环平稳可变核函数TF-MUSIC(CVCFTF-MUSIC)算法。该算法可显著提高DOA 估计的精确度，数值仿真验证了上述观点。

Direct Sequence/Frequency Hopping(DS/FH) signal composed by Direct Sequence Spread Spectrum(DSSS) and Frequency Hopping Spread Spectrum(FHSS) decreases the power density dramatically on time-frequency plane, and increases the estimation error of Direction-Of-Arrival(DOA) when classic TF-MUSIC(Time Frequency-Multiple Signal Classification) algorithm is applied in this circumstance. A novel scheme modified from classic TF-MUSIC algorithm is developed to resolve the problem. Firstly, the variable core function is designed on the parameters which character the DS/FH signal on time-frequency plane. Secondly, the cyclostationary frequency of DS/FH signal is estimated to construct the corresponding spatial covariance matrix. The modified algorithm is called Cyclostationary Variable Core Function TF-MUSIC(CVCFTF-MUSIC) algorithm. Simulation results validate that the proposed algorithm can increase the estimation precision of DOA estimation.