中文摘要：

笔迹性别识别在取证分析中具有重要意义。近年来,虽然笔迹性别识别获得了越来越多的关注,但是目前提出的算法都基于人工设计的特征,难以准确地表达笔迹包含的信息,因而准确率较低。针对这个问题,本文提出了一种基于深度学习的笔迹性别识别方法,使用深度学习caffe工具,将预处理后的笔迹图像输入本文设计的卷积神经网络进行分类。本文首先提取笔迹图像的每个单词,然后取单词的不同全排列拼接成基础图,接着按照固定的大小从基础图截取材料图,最后以材料图为输入数据,以包含7个卷积层的网络为模型进行分类。本文的方法在IAM On-Line公开数据库上进行了测试,取得了较高的识别率。

英文摘要：

Gender identification from handwriting is significant in forensic analysis. Although gender identification from handwriting gains more and more attention in recent years,all of existing algorithms are based on artificial features which can＇t precisely convey the information of handwriting,causing low accuracy. To solve the problem,this paper presents a method based on deep learning. With the tool of deep learning called caffe,the paper classifies preconditioned handwriting pictures by the proposed CNN. Firstly,extract words of a handwriting picture; then,combine different permutations of these words as basic picture and cut out stuff pictures of fixed size from the basic picture. Finally,stuff pictures are entered into network model containing seven convolutional layers to classify. The proposed method is evaluated on IAM On-Line database,and has achieved higher accuracy.