一种引入自适应动量项的变步长混沌信号盲分离算法

Variable-step Blind Source Separation Algorithm with Adaptive Momentum Item for Chaotic Signals

该文针对混沌信号盲分离问题,提出一种改进盲分离算法。该算法利用信号分离评价指标来构造函数实现步长和动量因子的自适应调整,然后将构造函数代入盲分离算法中并引入自适应动量项。区别于大多数算法不对混合矩阵进行估计的问题,该算法用变步长函数迭代估计出混合矩阵,从而得到全局矩阵和估计评价指标,以此迭代更新步长和动量因子,最终得出分离矩阵。仿真表明,该算法依据估计评价指标构造函数调整步长和动量因子方法是有效的,在平稳和非平稳环境下对混合混沌信号分离时都能达到收敛速度快且稳态误差小的效果;在混入色噪声时,比传统算法抗噪性能好,表明该文算法在混沌信号盲分离处理中有一定应用价值。

To solve the problem of blind source separation for chaotic signals, an improved blind separation algorithm is proposed. A function is constructed by signal separation evaluation index, which adaptively updates the step size and momentum factor, then substitutes the obtained variable step-size function into blind source separation algorithm and introduces the adaptive momentum item. Different from most algorithms which can not estimate the mixing matrix, the proposed algorithm estimates iteratively the mixing matrix by the variable step function, then the global matrix and the estimated evaluation can be obtained on which step and momentum factor are iteratively updated. Finally, the separation matrix is obtained. Simulations show that the algorithm is effective to adjust the step and momentum factor based on the estimated evaluation index constructor. In stationary and non-stationary environments, the algorithm has faster convergence speed and lower steady error for separating the mixed chaotic signals. When mixing color noise, the proposed algorithm is better than that of the traditional algorithm, which shows that the proposed algorithm has certain application value to the chaotic signal blind source separation processing.