论文部分内容阅读
为了更好地解决传统梯度下降算法中收敛点难以确定的难题,根据数字图像信号有界的特点,提出一种改进的梯度自适应在线独立分量分析(ICA)算法.该算法将传统梯度在线算法中的收敛点强加于学习过程,使其由传统的梯度下降过程变为上升过程,保证接收端在最后一组信号到达时,分离矩阵可保持在最优分离点上.理论分析和仿真结果表明,本算法具有较好的稳态性能和数值稳定性,是一种有效的ICA算法.
In order to solve the problem of difficult to determine the convergence point in the traditional gradient descent algorithm, an improved Gradient Adaptive Online Independent Component Analysis (ICA) algorithm is proposed according to the bounded characteristics of the digital image signal. The proposed algorithm uses the traditional gradient online algorithm The convergence point in the learning process is forced to change from the traditional gradient descent process to the ascending process to ensure that the separation matrix can be kept at the optimal separation point when the last group of signals arrives at the receiving end.A theoretical analysis and simulation results show that , This algorithm has good steady-state performance and numerical stability, is an effective ICA algorithm.