摘要
针对传统手写体数字识别精度不高的问题,本文设计了一种基于脉冲神经神经网络的手写体数字识别算法。该算法首先对输入的手写体数字进行预处理,然后提取手写体数字的二值图像轮廓和笔画次序等结构特征,最后根据脉冲神经网络模型分类,预测测试集合样本。该算法在MNIST字体库里的数据进行实验,精度可达98.7%,优于其它识别算法。
Aiming at the problem of traditional handwritten digit recognition accuracy is not high,a handwritten digit recognition based on spiking neural network algorithm is proposed in this paper.The handwritten digit which is inputted should be preprocessed firstly and then be extracted the binary image contour and stroke order. Finally,according to the classification of spiking neural network model,the sample size of test set is predicted.The algorithm is tested on the data of MNIST font library,and the accuracy is up to 98.7%,which is better than other recognition algorithms.
引文
[1]朱书眉.基于卷积神经网络的手写体数字识别[J].数字技术与应用,2018,36(11):44-46.
[2] Shruti R. Kulkarni. Spiking neural networks for handwritten digit recognition—Supervised learning and network optimization[J]. Neural networks,2018(103):118-127.
[3]高诺.采用卷积神经网络的数字手写体识别的研究[J].齐鲁工业大学学报,2018,32(5)45-49.
[4]马媛媛.卷积神经网络及其在手写体数字识别中的应用[J].湖北工程学院学报,2017,37(6):66-72.
[5]辛平.三相电流型PWM整流器的能量成型控制[J].煤矿机械,2016,37(8):132-134.
[6]刘德君.基于能量成型的双足机器人解耦控制[J].控制与决策,2012,27(12):1890-1892.
[7]曲萍萍.一种有效的组合电路测试生成算法[J].机床与液压,2018,46(18):100-103.
[8]周振雄.精密六自由度平面悬浮平台悬浮位置控制[J].机床与液压,2018,46(24):125-127.