首页> 外国专利> METHOD AND APPARATUS FOR RECOGNISING HANDWRITTEN CHARACTERS IN AN OPTICAL CHARACTER RECOGNITION MACHINE

METHOD AND APPARATUS FOR RECOGNISING HANDWRITTEN CHARACTERS IN AN OPTICAL CHARACTER RECOGNITION MACHINE

机译:在光学字符识别机中识别手写字符的方法和装置

摘要

1454813 Character recognition systems OPTICAL BUSINESS MACHINES Inc 6 Feb 1974 [25 July 1973] 05535/74 Heading G4R An array of binary signals representing a scanned character in a field is normalized by combining bits in pairs of rows and columns to produce a normalized array of bits representing a reduced size character. The pairs are selected in accordance with the degree of reduction necessary, as determined by the height and width of the character. Signals representing a scanned field containing a handwritten character are stored as an array of bits in 32 columns of 64-bit shift registers. Circuitry determines the position, height and width of the character by developing signals corresponding to the horizontal and vertical projections of the character as the signals are written serially into the registers. For example when a series of 64 bits, corresponding to a vertical slice of the field, contains one or more black (character-portion) bits a 1-bit is set in a width shift register. If the register contains 5 or more consecutive black bits to the right of a white bit the left hand edge of the character is assumed to have just entered the store. The position of the right hand edge of the character, as indicated by the right hand end of the series of black bits, indicates the character width and is used to select the degree of reduction needed. As indicated in Fig. 67, if the character is 15 bits wide no reduction is needed. If it is 19 bits wide a 5 : 4 reduction is needed and is obtained by combining the bits in corresponding rows of columns 4, 5; 9,10; 14, 15; 18, 19 of the store. Similar combination takes place for other reduction ratios. In a generally similar manner _the height of the character is obtained. To ensure that voids or specks do not adversely effect the height determination the determination is inhibited until 5 successive columns con- .taining blacks have been detected and is maintained until 2 successive columns containing no blacks have been detected. The pairs of rows to be combined to reduce the character height are as indicated in Fig. 68. The signals representing the character thus normalized are stored in 15 columns of 24-bit registers. In addition the normalized character is shifted right to centralize it in the normalized store. Electronic masking is used to detect 7 line features of the character. One such mask is 15 bits wide by 4 bits high. If a black bit occurs in any of the four mask rows and across 8 consecutive columns the presence of a horizontal line is indicated. The mask is scanned vertically down the store character to detect such a line in the top, middle, and bottom regions of the store. Other masks, each comprising a number of sub-masks (Figs. 39-66, not shown) detect lines in the upper and lower left and right portions of the store. The 7-bit signal from the masks passes to recognition circuitry comprising gates which detect particular combinations of mask outputs to identify the character or to indicate an error in the case of unresolvable ambiguity. The recognition circuitry provides a signal on one of a number of leads corresponding to the various characters which can be recognized and feeds a coding unit which provides a corresponding parallel-bit signal to a utilization device.
机译:1454813字符识别系统OPTICAL BUSINESS MACHINES Inc. 1974年2月6日[1973年7月25日] 05535/74标题G4R通过组合成对的行和列中的位以产生一个归一化的数组,对表示字段中扫描字符的二进制信号阵列进行归一化代表缩小字符的位。根据字符的高度和宽度所确定的必要缩小程度来选择对。表示包含手写字符的已扫描字段的信号以位阵列的形式存储在32列64位移位寄存器中。当信号被串行写入寄存器时,电路通过产生与字符的水平和垂直投影相对应的信号来确定字符的位置,高度和宽度。例如,当对应于该字段的垂直切片的一系列64位包含一个或多个黑色(字符部分)位时,在宽度移位寄存器中设置1位。如果寄存器在白色位的右边包含5个或更多连续的黑色位,则假定该字符的左手边刚刚进入存储区。如一系列黑色位的右端所示,字符右边缘的位置指示字符宽度,并用于选择所需的缩小程度。如图67所示,如果字符是15位宽,则不需要缩小。如果它是19位宽,则需要5:4的缩减,这是通过将列4、5的相应行中的位组合而获得的; 9,10; 14、15; 18、19的商店。其他减速比也发生类似的组合。以通常类似的方式获得字符的高度。为了确保空洞或斑点不会对高度确定产生不利影响,在检测到5个连续的包含黑色的列之前,禁止进行该确定,并保持该状态直到检测到连续2个不包含黑色的列。为减少字符高度而要组合的成对的行如图68所示。代表这样标准化的字符的信号被存储在15列的24位寄存器中。另外,将规范化字符右移以将其集中在规范化存储中。电子遮罩用于检测字符的7个线条特征。一个这样的掩码是15位宽乘4位高。如果在四个掩码行中的任何一个中以及在连续的8个列中出现黑位,则表明存在水平线。在商店角色的垂直下方扫描遮罩,以检测商店顶部,中间和底部区域中的此类线条。其他每个都包含多个子遮罩的遮罩(图39-66,未显示)检测商店左上方和右下方的线条。来自掩码的7位信号传递到包含门的识别电路,这些门检测掩码输出的特定组合以识别字符或在无法解决的歧义情况下指示错误。识别电路在对应于可以被识别的各种字符的多个引线之一上提供信号,并将其提供给编码单元,该编码单元将相应的并行位信号提供给利用设备。

著录项

  • 公开/公告号FR2238971A1

    专利类型

  • 公开/公告日1975-02-21

    原文格式PDF

  • 申请/专利权人 OPTICAL BUSINESS MACHINES INCUS;

    申请/专利号FR19740008159

  • 发明设计人

    申请日1974-03-11

  • 分类号G06K9/18;

  • 国家 FR

  • 入库时间 2022-08-23 03:47:16

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号