论文部分内容阅读
噪声数字的识别具有很好的应用前景,也是后期处理的基础。基于离散Hopfield神经网络的联想记忆能力,通过改进神经网络的记忆样本,再利用Hebb规则对改进的记忆样本进行学习,得到权值矩阵,根据待识别的噪声数字的信息联想起记忆的数字。利用改进后的离散Hopfield神经网络对噪声数字进行了识别的实验。实验结果表明,该方法提高了传统网络的记忆能力和识别的正确率。
The identification of noise figures has good application prospects, but also the basis for post-processing. Based on the associative memory ability of discrete Hopfield neural network, we improve the memory samples of the neural network, and then use Hebb rules to study the improved memory samples to get the weight matrix, and associate the memory numbers according to the information of the noise numbers to be recognized. Experiments to identify noise figures using improved discrete Hopfield neural networks. Experimental results show that this method improves the memory capacity of traditional networks and the accuracy of recognition.