纯相位二值化匹配滤波器的优化设计

Optimized Design of Binary Phase-only Matched Filters and Their Optical Implementation

通过采用神经网络中的Clipping方法和MonteCarlo修改学习算法，对用于光学模式识别的纯相位二值化匹配滤波器进行了优化设计。计算机模拟结果表明，和传统的纯相位匹配滤波器的相关输出结果相比，其识别输出的信噪比和信号相关峰值得到了明显的提高，从而为今后的光学实现奠定了良好的基础。

An optimized design of binary phase-only matched filters used in optical pattern recognition by means of Clipping and Monte Carlo learning arithmetic in the neural network is introduced. The computer simulation result indicates that compared with the correlation output of the traditional phase-only matched filter, their signal-to-noise ratio (SNR) for recognizing output and the signal correlation peak are improved obviously.Thereby it may provide a favorable foundation for the optical implementation in the future.