CN104063705B - A method and apparatus for handwriting feature extraction - Google Patents

A method and apparatus for handwriting feature extraction Download PDF

Info

Publication number
CN104063705B
CN104063705B CN 201410247878 CN201410247878A CN104063705B CN 104063705 B CN104063705 B CN 104063705B CN 201410247878 CN201410247878 CN 201410247878 CN 201410247878 A CN201410247878 A CN 201410247878A CN 104063705 B CN104063705 B CN 104063705B
Authority
CN
Grant status
Grant
Patent type
Application number
CN 201410247878
Other languages
Chinese (zh)
Other versions
CN104063705A (en )
Inventor
曹骥
李健
张连毅
武卫东
Original Assignee
北京捷通华声语音技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Grant date

Links

Abstract

本申请提供了一种笔迹特征提取的方法和装置,包括:按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据;依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量;获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标;根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征;依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 The present application provides a method and apparatus for handwriting feature extraction, comprising: writing data in time series acquired in accordance with the stroke and pre-stroke data, stroke data obtained after pretreatment; according to the time sequence of the pre-processed interval segment handwriting data, etc., to obtain a plurality of stroke segments vectors; obtaining the plurality of online feature stroke segments vectors, wherein the line comprises a plurality of angles and the center coordinates of the stroke segments vectors; according to the pre- obtaining handwriting data processing center of gravity of the stroke data, according to the center of gravity of the offline feature extraction stroke data after the preprocessing; according to the features of online and offline features, numerical normalization, the said normalized value as a result of the stroke characteristic data acquisition. 因此,本申请解决了联机手写笔迹数据特征识别准确率低的问题。 Accordingly, the present application to solve the data-line handwritten character recognition problem of low accuracy.

Description

一种笔迹特征提取的方法和装置 A method and apparatus for handwriting feature extraction

技术领域 FIELD

[0001] 本申请涉及联机手写汉字识别技术领域,特别是涉及一种笔迹特征提取的方法和装置。 [0001] The present application relates to the field of online handwritten character recognition technology, particularly to a method and apparatus for handwriting feature extraction.

背景技术 Background technique

[0002] 联机手写的笔迹数据,由于书写者的书写习惯、笔迹采集设备的精度的不同,对于相同的文字,存在很大的形变和图形上的差异,因此,对于笔迹数据的特征提取提出了更高的要求,需要在笔迹数据特征提取时能够有效的表达手写笔迹数据的本质,体现相同字的笔迹数据的相同点,并区分不同字的笔迹数据的不同点。 [0002] The data-line handwritten stroke due writer writing habits, different precision handwriting capture devices, for the same word, there is a difference in the pattern and a large deformation, and therefore, the handwriting data for the feature extraction is proposed higher requirements need to effectively express the essence of handwriting data when handwriting data feature extraction, reflect the same point of the stroke data of the same word, different from stroke data and distinguish different words.

[0003] 目前传统笔迹数据的特征提取方法,是通过对笔迹数据进行顺序扫描方法和网格统计方法进行笔迹数据特征的提取。 [0003] It is characteristic of traditional handwriting data extraction method for extracting features of handwriting data by handwriting data grid sequential scanning method and statistical methods. 其中,笔迹数据进行顺序扫描方法是按照书写的顺序对笔迹数据中的点位置或角度进行扫描,该方法未考虑笔迹数据特征中的轨迹的角度变化信息,也未考虑笔迹数据特征相邻角度间的相似性。 Wherein the handwriting data is sequential scanning method or an angle position in the handwriting data written in the order of scanning, this method does not track the feature stroke data information into angle change, characterized in handwriting data without regard for the angle between adjacent the similarity. 网络统计方法是按照等宽和等高的方式进行笔迹数据特征的提取,该方法未考虑笔迹数据特征对称投影的信息,并且,该方法存在手写笔迹特征的提取过于机械、规整以及形变适应性不佳的问题。 Network Statistics handwriting data extraction method according to feature width and contour manner, this method does not consider the handwriting data characteristic symmetrical projection information, and the method to extract the presence of features of handwriting too mechanical, warping and deformation are not flexible good question.

[0004] 上述方法,存在笔迹数据特征提取不全面和笔迹数据特征适应性不佳的问题,以上问题严重影响了后续分类器的分类效果,进而导致了联机手写笔迹数据特征识别准确率低的问题。 [0004] The above-described method, there is handwriting data feature extraction incomplete and poor adaptability features of handwriting data problems, the above problems seriously affect the classification performance of classifier follow, leading to the data-line handwritten character recognition problem of low accuracy .

发明内容 SUMMARY

[0005] 本申请提供一种笔迹特征提取的方法和装置,以解决联机手写笔迹数据特征识别准确率低的问题。 [0005] The present application provides a method and apparatus for handwriting feature extraction, data to address the on-line handwritten character recognition problem of low accuracy.

[0006] 为了解决上述问题,本申请公开了一种笔迹特征提取的方法,包括: [0006] In order to solve the above problems, the present application discloses a method of handwriting feature extraction, comprising:

[0007] 按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据; [0007] handwriting data written in accordance with the time-series data acquisition and pre-stroke, stroke data obtained after pretreatment;

[0008] 依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量; [0008] The time series based on handwriting data segment intervals for the pretreatment, to obtain a plurality of stroke segments vectors;

[0009] 获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标; [0009] obtaining the plurality of online feature stroke segments vectors, wherein the connection comprises a plurality of angles and a center of the stroke segments vectors coordinates;

[0010] 根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征; [0010] The center of gravity of the acquired data based on the handwriting stroke data after the pretreatment, the offline feature extraction stroke data after the pretreatment according to the gravity;

[0011] 依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 [0011] according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition.

[0012] 优选地,所述按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据的步骤包括: [0012] Preferably, the writing time series data acquired in accordance with the handwriting data and pre-stroke, after the step of obtaining pre-stroke data comprises:

[0013] 将采集的笔迹数据按照书写的时间序列进行线性尺寸规整化后,获得各个自然笔画段长度; After [0013] The stroke data acquisition is performed according to the writing of the linear size regular time series, to obtain the respective stroke length NATURAL;

[0014] 依据获得的所述各个自然笔画段长度,得到由所述各个自然笔画段组成的笔迹数据的长度。 [0014] is obtained according to the nature of each stroke length, stroke data obtained by the length of the respective segments of the natural stroke.

[0015] 优选地,所述多个笔画段向量的角度包括:各个笔画段向量与X轴正方向的角度、 各个笔画段向量与Y轴正方向的角度以及各个笔画段向量与其相邻的笔画段向量之间的角度。 The angle [0015] Preferably, the plurality of stroke segments vectors comprising: an angle vector and the positive direction of the X-axis of each stroke segments, each angular stroke segments vectors and Y-axis and the positive direction of the respective adjacent stroke segments vectors stroke the angle between the segments vectors.

[0016] 优选地,其特征在于,所述脱机特征包括投影脱机特征、或网格脱机特征、或扇形脱机特征、或轮廓脱机特征。 [0016] Preferably, wherein said feature comprises a projection offline offline features, mesh or offline features, or fan-off features, characteristics or profile offline.

[0017] 优选地,当所述脱机特征为所述投影脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的脱机特征的步骤包括: [0017] Preferably, when the offline feature wherein said projection offline, the offline feature extraction after the pretreatment of handwriting data based on said center of gravity comprises the step of:

[0018] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上部分区域和下部分区域,从垂直方向分割为左部分区域和右部分区域,分别扫描各个笔画段向量的中心坐标在所述上部分区域、下部分区域、左部分区域和右部分区域出现的个数; [0018] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data divided from the horizontal direction to an upper portion and a lower region partial regions divided from a vertical direction to the left area and the right partial region portion, respectively, the scanning center coordinates of each stroke segments vectors in the upper partial region, the number of lower partial region, a left region and a right portion of the partial region appears;

[0019] 当所述脱机特征为所述网格脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的脱机特征的步骤包括: Step [0019] When the offline feature offline wherein said mesh, according to the center of gravity of the offline feature extraction after the pretreatment stroke data comprises:

[0020] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0020] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0021] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上网格和下网格,从垂直方向分割为左网格和右网格,分别扫描各个笔画段向量的中心坐标在所述上网格、下网格、 左网格和右网格的八个方向上出现的个数; [0021] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, the stroke data is divided into a grid on the pretreatment and the net horizontal direction cells divided from a vertical direction to the left and right mesh grids, respectively the scanning center coordinates of each stroke segments vectors number appearing on the eight directions in the grid, grid lower, left and right grid mesh ;

[0022] 当所述脱机特征为所述扇形脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的扇形脱机特征的步骤包括: [0022] wherein when the step of the offline feature when the fan off, according to the center of gravity of the offline feature extraction fan preprocessed stroke data comprises:

[0023] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0023] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0024] 以所述笔迹数据的重心为圆心,对所述预处理后的笔迹数据分割为多个扇形区域,分别扫描各个笔画段向量的中心坐标在八个方向上出现的个数; [0024] In the center of gravity as the center of the stroke data, handwriting data after the pretreatment is divided into a plurality of fan-shaped region, respectively, each of the scanning center coordinates of the stroke segments vectors number appearing in eight directions;

[0025] 当所述脱机特征为所述轮廓脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的轮廓脱机特征的步骤包括: Step [0025] When the off-line profile wherein the offline feature, the center of gravity based on the extracted outline feature offline after the pretreatment stroke data comprises:

[0026] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0026] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0027] 以所述笔迹数据的重心为结束点,分别扫描各个笔画段向量的中心坐标在八个方向出现的个数。 [0027] In the center of gravity of the stroke data is an end point, respectively, the scanning center coordinates of each stroke segments vectors occurring in the number of eight directions.

[0028] 为了解决上述问题,本申请还公开了一种笔迹特征提取的装置,包括: [0028] In order to solve the above problems, the present application also discloses a device for handwriting feature extraction, comprising:

[0029] 获取模块,用于按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据; [0029] The acquisition module configured in accordance with the handwriting data written in the time-series data acquisition and pre-stroke, stroke data obtained after pretreatment;

[0030] 分割模块,用于依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段, 获得多个笔画段向量; [0030] The segmentation module, a time sequence according to the stroke of the pre-processed data segment intervals to obtain a plurality of stroke segments vectors;

[0031] 计算模块,用于获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标; [0031] calculation means for obtaining the plurality of online feature stroke segments vectors, wherein the line comprises a plurality of angles and a center of the stroke segments vectors coordinates;

[0032] 提取模块,用于根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征; [0032] extraction means for acquiring the data based on the handwriting stroke data after the pretreatment center of gravity, the offline feature extraction stroke data after the pretreatment according to the gravity;

[0033] 处理模块,用于依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 [0033] The processing module configured according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition.

[0034] 优选地,所述获取模块包括:线性规整模块,用于将采集的笔迹数据按照书写的时间序列进行线性尺寸规整化后,获得各个自然笔画段长度; [0034] Preferably, the obtaining module comprises: a linear structured module for the handwriting data collected after regular linear dimensions, to thereby obtain the respective stroke length in the writing of natural time series;

[0035] 长度获取模块,用于依据获得的所述各个自然笔画段长度,得到由所述各个自然笔画段组成的笔迹数据的长度。 [0035] The acquisition module length, stroke length for each segment of the natural basis obtained, the length of stroke data obtained by the respective segments of the natural stroke.

[0036] 优选地,所述多个笔画段向量的角度包括:各个笔画段向量与X轴正方向的角度、 各个笔画段向量与Y轴正方向的角度以及各个笔画段向量与其相邻的笔画段向量之间的角度。 The angle [0036] Preferably, the plurality of stroke segments vectors comprising: an angle vector and the positive direction of the X-axis of each stroke segments, each angular stroke segments vectors and Y-axis and the positive direction of the respective adjacent stroke segments vectors stroke the angle between the segments vectors.

[0037] 优选地,所述脱机特征包括投影脱机特征、或网格脱机特征、或扇形脱机特征、或轮廓脱机特征。 [0037] Preferably, the offline feature comprises a projection offline features, mesh or offline features, or fan-off features, characteristics or profile offline.

[0038] 优选地,当所述脱机特征为所述投影脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的脱机特征时: [0038] Preferably, when the offline wherein said projection offline feature, the feature extraction module extracts offline after the pretreatment according to the handwriting data in the center of gravity:

[0039] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上部分区域和下部分区域,从垂直方向分割为左部分区域和右部分区域,分别扫描各个笔画段向量的中心坐标在所述上部分区域、下部分区域、左部分区域和右部分区域出现的个数; [0039] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data divided from the horizontal direction to an upper portion and a lower region partial regions divided from a vertical direction to the left area and the right partial region portion, respectively, the scanning center coordinates of each stroke segments vectors in the upper partial region, the number of lower partial region, a left region and a right portion of the partial region appears;

[0040] 当所述脱机特征为所述网格脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的脱机特征时: [0040] When the offline feature wherein said mesh offline, the offline feature extraction module extracts the stroke data after the pretreatment in accordance with the center of gravity:

[0041] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0041] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0042] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上网格和下网格,从垂直方向分割为左网格和右网格,分别扫描各个笔画段向量的中心坐标在所述上网格、下网格、 左网格和右网格的八个方向上出现的个数; [0042] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, the stroke data is divided into a grid on the pretreatment and the net horizontal direction cells divided from a vertical direction to the left and right mesh grids, respectively the scanning center coordinates of each stroke segments vectors number appearing on the eight directions in the grid, grid lower, left and right grid mesh ;

[0043] 当所述脱机特征为所述扇形脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的扇形脱机特征时: [0043] When the feature is off when the fan off feature, the feature extraction module extracts the fan off after the pretreatment according to the handwriting data in the center of gravity:

[0044] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0044] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0045] 以所述笔迹数据的重心为圆心,对所述预处理后的笔迹数据分割为多个扇形区域,分别扫描各个笔画段向量的中心坐标在八个方向上出现的个数; [0045] In the center of gravity as the center of the stroke data, handwriting data after the pretreatment is divided into a plurality of fan-shaped region, respectively, each of the scanning center coordinates of the stroke segments vectors number appearing in eight directions;

[0046] 当所述脱机特征为所述轮廓脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的轮廓脱机特征时: [0046] When the off-line profile wherein the offline feature, the feature extraction module extracts the outline of offline handwriting data after the pretreatment in accordance with the center of gravity:

[0047] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0047] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0048] 以所述笔迹数据的重心为结束点,分别扫描各个笔画段向量的中心坐标在八个方向出现的个数。 [0048] In the center of gravity of the stroke data is an end point, respectively, the scanning center coordinates of each stroke segments vectors occurring in the number of eight directions.

[0049] 与现有技术相比,本申请包括以下优点: [0049] Compared with the prior art, the present application includes the following advantages:

[0050] 首先,本申请依据时间序列对预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量的联机特征,所述联机特征包括多个笔画段向量的角度和中心坐标。 [0050] First, the present application for handwriting data pre-processed and the like based on the time sequence of the spacer segment, a plurality of stroke segments vectors online feature, the online feature comprises a plurality of angle and center coordinates of the stroke segments vectors. 通过计算多个笔画段向量的角度和中心坐标,从而使笔迹数据的特征提取覆盖了笔迹数据的局部特性和全局特性,避免了现有方法中仅考虑笔迹数据特征点的位置,从而造成笔迹数据特征提取不全面的问题。 By calculating the center coordinates and the angle of the plurality of stroke segments vectors, the stroke data so that the extracted feature local coverage characteristics and global characteristics of handwriting data, existing techniques only consider the position of the feature point handwriting data, the handwriting data resulting feature extraction incomplete question.

[0051] 其次,本申请通过对预处理后的笔迹数据获取笔迹数据的重心,并依据重心进行对称投影,然后提取相邻区域的笔迹数据的局部特征和全局特性,从而避免了等宽和等高方式进行提取笔迹数据特征时过于机械和形变适应性不佳的问题。 [0051] Next, the present application by pretreatment of handwriting data acquisition center of gravity of the handwriting data, based on projected center of gravity and symmetrically and extracts local features and global characteristics of the regions adjacent to the handwriting data, thus avoiding the width and the like when handwriting data feature extraction methods are too high mechanical strain and poor adaptability.

[0052] 再次,本申请通过对提取的联机特征和脱机特征的组合,得到了有效的笔迹数据特征,进而保证了后续分类器训练的可靠性,并显著提高了分类器的分类准确度,最终提高了联机手写的识别准确率。 [0052] Again, the present application features through a combination of online and offline feature extraction, has been effective feature stroke data, thereby ensuring the reliability of a subsequent classifier training, and significantly improves the classification accuracy of the classifier, and ultimately improve the accuracy of online handwriting recognition.

附图说明 BRIEF DESCRIPTION

[0053] 图1是本申请实施例一中的一种笔迹特征提取的方法的流程图; [0053] FIG. 1 is a flowchart of a method of the present application embodiment a handwriting feature extraction in the embodiment;

[0054] 图2是本申请中的笔迹数据经采集设备后采集的笔迹数据示意图; [0054] FIG. 2 is a schematic diagram of the handwriting data in the handwriting data of the present application after the collecting devices;

[0055] 图3是本申请实施例二中的一种笔迹特征提取的方法的流程图; [0055] FIG. 3 is a flowchart of a method of the present application handwriting characteristics of two embodiments of the extracting embodiment;

[0056] 图4是本申请中的笔画段向量与相邻笔画段向量的夹角示意图; [0056] FIG. 4 is a schematic view of the angle between stroke segments vectors stroke segments vectors neighboring the present application;

[0057] 图5是本申请中的以笔迹数据的重心为分割点的投影脱机特征的示意图; [0057] FIG. 5 is the center of gravity in the present application is divided stroke data is projected schematic offline feature point;

[0058] 图6是本申请中的二维平面的八个方向的示意图; [0058] FIG. 6 is a schematic diagram of the two-dimensional plane of the eight directions in the present application;

[0059] 图7是本申请中的以笔迹数据的重心为分割点的扇形脱机特征的示意图; [0059] FIG. 7 is to the center of gravity in the present application is divided stroke data sector schematic offline feature points;

[0060] 图8是本申请实施例三中的一种笔迹特征提取装置的结构框图。 [0060] FIG. 8 is a handwriting feature block diagram in Example III extracting apparatus embodiment of the present application.

具体实施方式 detailed description

[0061] 为使本申请的上述目的、特征和优点能够更加明显易懂,下面结合附图和具体实施方式对本申请作进一步详细的说明。 [0061] The above object of the present application, features and advantages can be more fully understood in conjunction with the accompanying drawings and the following specific embodiments of the present application will be further described in detail.

[0062] 参照图1,示出了本申请实施例一中的一种笔迹特征提取的方法,包括: [0062] Referring to Figure 1, there is shown a method of the present application embodiment a handwriting feature extraction implementation, comprising:

[0063] 步骤101:按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据。 [0063] Step 101: according to the handwriting data written in the time-series data acquisition and pre-stroke, stroke data obtained after pretreatment.

[0064] 其中,笔迹数据书写的时间序列是通过采集设备获取的。 [0064] where handwriting data written in time series acquired by the acquisition device.

[0065] —种笔迹数据经采集设备后采集的笔迹数据如图2所示,其中,笔迹数据经过采集设备后,采集到一系列的数据坐标点,数据坐标点包括每个点的横坐标值和纵坐标值,以及,每个笔画的结束标志和整个字的结束标志。 [0065] - stroke after stroke data kind of data collecting devices shown in Figure 2, wherein, after the handwriting data acquisition device to collect a series of coordinate point data, coordinate data of the point values ​​for each point of the abscissa including ordinate values ​​and end flag, and a flag end of each stroke and the entire word. 例如:采集到的数据坐标点包括(Xo,Yo)、 (X^Y1)、(X2,Y2)……(Xn,Yn)。 For example: the collected data comprises coordinate point (Xo, Yo), (X ^ Y1), (X2, Y2) ...... (Xn, Yn). 采集到的一系列数据坐标点中包含有笔迹数据的本质特征, 可以依据这些特征对笔迹数据进行处理,进而提取笔迹特征。 Data collected from a series of coordinate points included in the stroke has the essential characteristics of data, handwriting data may be processed according to these features, then extract handwriting features.

[0066] 步骤102:依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量。 [0066] Step 102: handwriting data segment intervals for the pretreatment according to the time sequence to obtain a plurality of stroke segments vectors.

[0067] 根据采集设备对用户书写的时间序列的笔划进行准确的等间隔分段,分段后的笔迹数据即为笔画段向量。 [0067] accurate intervals segment collection device according to the stroke of the user writing time series, the segmented stroke data is the stroke segments vectors.

[0068] 步骤103:获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标。 [0068] Step 103: obtaining the plurality of online feature stroke segments vectors, wherein said line includes an angle of the plurality of stroke segments vectors, and the center coordinates.

[0069] 笔画段向量的中心坐标可以通过以下公式获得: [0069] The center coordinate of the stroke segments vectors may be obtained by the following equation:

[0070] [0070]

Figure CN104063705BD00091

[0071] 其中,Xi为笔画段向量的起始坐标,Xi+i为笔画段向量的中止坐标。 Suspension coordinate [0071] wherein Xi is the initial coordinate stroke segments vectors, Xi + i for the stroke segments vectors.

[0072] 步骤104:根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征。 [0072] Step 104: acquiring the center of gravity of the handwriting data based on the handwriting data of the pretreatment, the offline feature extraction stroke data after the pretreatment according to the center of gravity.

[0073] 需要说明的是,上述步骤103和104在实际执行时并不限于上述顺序,也可以步骤104在步骤103之前执行,还可以二者并行执行。 [0073] Incidentally, the above-described steps 103 and 104 in the actual implementation is not limited to the above sequence, step 104 may be performed before step 103, both may also be performed in parallel.

[0074] 步骤105:依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 [0074] Step 105: according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition.

[0075] 其中,数值归一化处理的结果的范围可以由本领域技术人员根据实际情况适当设定,优选为0-8。 [0075] wherein the value of the result of the normalization process can be appropriately set range by those skilled in the art based on the actual situation, preferably 0-8.

[0076] 特征是指某一物质自身所具备的特殊性质,是区别于其他物质的基本征象和标志。 [0076] wherein a refers to the special nature of the material itself have, and the flag is different from the basic signs of other substances. 对于联机手写的笔迹特征,是指笔迹书写方式和形状上特性。 For online handwritten handwriting characteristics, handwriting writing means and the characteristic mode shape.

[0077] 将数值归一化处理的结果作为采集的笔迹数据的特征可以通过以下方式进行文字的识别。 [0077] The normalized value as a result of feature stroke data collected can be recognized characters in the following manner.

[0078] 首先,将提取的笔迹数据的特征与字库的模板比较,将与提取的笔迹数据的特征匹配率大的字列出,供使用者选择,使用者选择正确的输入文字后,即完成手写文字的识别。 [0078] First, the comparison of the extracted feature character with stroke data template, the rate of the data feature matching with the extracted words are listed in a large stroke for the user to select the user select the correct input characters, the complete recognize handwritten text.

[0079] 其中,字库的模板的建立过程包括:对字典中已知的文字,由训练者手写输入,建立字典中文字与手写文字的对应关系,训练者手写输入的文字作为已知文字的模板。 [0079] wherein the process of establishing character template comprising: a dictionary of known words, the handwriting input by the trainer, establishing the corresponding relationship dictionary text and handwritten characters, handwritten input character trainer known as a template text . 同样的文字可以由多个训练者手写输入,重复多次,从而一个文字可以对应多个手写模板。 The same character may be input by a plurality of handwritten trainer, repeated several times, so that a handwritten character may correspond to a plurality of templates. 在匹配时,可以将手写输入的文字与多个文字的多个模板匹配。 In a match, you can match multiple templates with multiple text handwriting input text.

[0080] 需要说明的是,本申请只列举了一种将提取的笔迹数据的特征进行文字识别的方法,可以采用现有技术中任何方式对所提取的笔迹数据的特征进行文字识别,本申请不加以限制 [0080] Incidentally, the present application features include only the extracted stroke data A character recognition method is performed, the extracted features may handwriting data subjected to character recognition using any of the prior art, the present application not be limited

[0081] 通过本实施例,首先,本申请依据时间序列对预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量的联机特征,所述联机特征包括多个笔画段向量的角度和中心坐标。 [0081] By the present embodiment, firstly, the present application the handwriting data were based on pre-segmented time series at equal intervals, a plurality of line feature stroke segments vectors, wherein the line comprises a plurality of angular stroke segments vectors and center coordinates. 通过计算多个笔画段向量的角度和中心坐标,从而使笔迹数据的特征提取覆盖了笔迹数据的局部特性和全局特性,避免了现有方法中仅考虑笔迹数据特征点的位置,从而造成笔迹数据特征提取不全面的问题。 By calculating the center coordinates and the angle of the plurality of stroke segments vectors, the stroke data so that the extracted feature local coverage characteristics and global characteristics of handwriting data, existing techniques only consider the position of the feature point handwriting data, the handwriting data resulting feature extraction incomplete question.

[0082] 其次,本申请通过对预处理后的笔迹数据获取笔迹数据的重心,并依据重心进行对称投影,然后提取相邻区域的笔迹数据的局部特征和全局特性,从而避免了等宽和等高方式进行笔迹特征的提取时过于机械和形变适应性不佳的问题。 [0082] Next, the present application by pretreatment of handwriting data acquisition center of gravity of the handwriting data, based on projected center of gravity and symmetrically and extracts local features and global characteristics of the regions adjacent to the handwriting data, thus avoiding the width and the like conduct handwriting feature extraction way too high mechanical deformation and poor adaptability.

[0083] 再次,本申请通过对提取的联机特征和脱机特征的组合,得到了有效的笔迹数据特征,进而保证了后续分类器训练的可靠性,并显著提高了分类器的分类准确度,最终提高了联机手写的识别准确率。 [0083] Again, the present application features through a combination of online and offline feature extraction, has been effective feature stroke data, thereby ensuring the reliability of a subsequent classifier training, and significantly improves the classification accuracy of the classifier, and ultimately improve the accuracy of online handwriting recognition.

[0084] 参照图3,示出了本申请实施例二中的一种笔迹特征提取的方法。 [0084] Referring to FIG. 3, the present application shows a method of handwriting characteristics of two embodiments of the extracting embodiment.

[0085] 本实施例中,一种笔迹特征提取的方法,包括: [0085] In this embodiment, the feature extraction method of handwriting, comprising:

[0086] 步骤301:按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据。 [0086] Step 301: according to the handwriting data written in the time-series data acquisition and pre-stroke, stroke data obtained after pretreatment.

[0087] 本实施例中,通过采集设备采集到笔迹数据的一系列坐标点。 [0087] In this embodiment, a series of coordinate points acquired by the data acquisition device stroke. 其中,坐标点包括每个坐标点的横坐标值和纵坐标值,以及每个笔画的起点坐标,每个笔画的中止坐标,每个笔画的结束坐标和整个字的结束坐标。 Wherein the coordinate point values ​​comprising abscissa and ordinate values, and the coordinates of the starting point of each stroke each coordinate point, the coordinates of the end of each stroke suspended coordinates, and the coordinates of the end of each stroke of the entire word.

[0088] 在采集到笔迹数据后,按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据。 [0088] After sampling the handwriting data, handwriting data written in accordance with the time-series data acquisition and pre-stroke, stroke data obtained after pretreatment.

[0089] 步骤302 :将采集的笔迹数据按照书写的时间序列进行线性尺寸规整化,规整到96*96的尺寸,然后获得各个自然笔画段长度。 [0089] Step 302: the linear stroke data collection of regular size, the size of 96 * 96 structured to then obtain respective NATURAL stroke length in the writing time series.

[0090] 线性尺寸规整化是指通过伸缩变换统一采集的笔迹数据的尺寸,采用旋转、平移等变换改变采集的笔迹数据的位置。 [0090] Regularization linear dimension refers to the size scaled by a unified collection of handwriting data, using position of the rotation, translation and other transformation changes the stroke data acquisition.

[0091] 自然笔画段是指用户在书写过程中的横、竖、撇、捺。 [0091] NATURAL stroke cross-section refers to the user during writing in, vertical, left, right,.

[0092] 通过以下公式获得各个自然笔画段长度: [0092] Natural respective stroke length obtained by the following equation:

[0093] [0093]

Figure CN104063705BD00101

[0094] 其中,Xi为自然笔画段的起点横坐标值,Xi+i为自然笔画段的中止横坐标值,Yi为自然笔画段的起点纵坐标值,Y1+1为自然笔画段的中止坐标的纵坐标值。 [0094] wherein, Xi is the starting point of the abscissa value NATURAL stroke segment, Xi + i is the abscissa value of the suspension stroke segment of the natural, Yi is the starting ordinate values ​​of the natural stroke segment, Y1 + 1 is a stroke segment of the natural coordinate abort the ordinate values.

[0095] 依据⑴式可以获得各个自然笔画段长度,依据获得的各个自然笔画段长度,得到由各个自然笔画段组成的笔迹数据的长度。 [0095] The formula can be obtained based on the respective natural ⑴ stroke length, based on the respective stroke length naturally obtained, the length of stroke data obtained from the respective natural stroke segments.

[0096] 可以通过以下公式获得笔迹数据的长度: [0096] the length of the handwriting data may be obtained by the following equation:

[0097] [0097]

Figure CN104063705BD00102

[0098] 其中,η为点的数目,I1为自然笔画段长度。 [0098] wherein, η is the number of points, I1 natural stroke length.

[0099] 需要说明的是,线性尺寸规整化的范围可以由本领域技术人员根据实际情况适当设定,优选为规整到96*96的尺寸。 [0099] Incidentally, the linear dimensions of the regular range can be appropriately determined by those skilled in the art based on the actual situation, preferably structured to size 96 * 96.

[0100] 步骤303:依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量; [0100] Step 303: perform handwriting data segment intervals after the pretreatment according to the time sequence to obtain a plurality of stroke segments vectors;

[0101] 可以通过以下公式获得笔画段向量: [0101] stroke segments vectors may be obtained by the following equation:

[0102] [0102]

Figure CN104063705BD00103

[0103] 其中,Iv为笔画段向量,nd为特征向量的维数,nd可以任意取值,只要可以等分笔迹数据的长度即可。 [0103] where, Iv is the stroke segments vectors, nd is the number of dimensions of feature vectors, nd be any value, as long as the length of the handwriting data can aliquots.

[0104] 步骤304:获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标;所述多个笔画段向量的角度包括:各个笔画段向量与X轴正方向的角度、各个笔画段向量与Y轴正方向的角度以及各个笔画段向量与其相邻的笔画段向量之间的角度。 [0104] Step 304: obtaining the plurality of online feature stroke segments vectors, wherein the line comprises a plurality of angle and center coordinates of the stroke segments vectors; angles of the plurality of stroke segments vectors include: individual stroke segments vector angle X-axis positive direction, the angle between the respective angular stroke segments vectors and Y-axis and the positive direction of the respective adjacent stroke segments vectors stroke segments vectors.

[0105] 其中,各个笔画段向量与X轴正方向的角度范围和各个笔画段向量与Y轴正方向的角度范围是0-180度。 [0105] wherein the vector and the positive angular range in the X axis direction and an angular extent of each stroke segment vector and the positive direction of the Y-axis of each stroke segment is 0-180 degrees.

[0106] —种笔画段向量与相邻笔画段向量的夹角如图4所示。 [0106] - Species stroke segments vectors and the angle between adjacent stroke segments vectors as shown in FIG. 其中,笔画段向量0与笔画段向量1是相邻向量,笔画段向量间的夹角由笔画段向量O与笔画段向量1计算获得的,笔画段向量与其相邻的笔画段向量之间的角度的范围为0-180度。 Wherein the stroke segments vectors 0 and 1 are adjacent stroke segments vectors vector, the angle between the stroke segments vectors stroke segments vectors of the stroke segments vectors O 1 obtained by calculation, between stroke segments vectors adjacent stroke segments vectors angle range of 0-180 degrees.

[0107] 步骤305:根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征; [0107] Step 305: acquiring the center of gravity of the handwriting data based on the handwriting data of the pretreatment, the offline feature extraction stroke data after the pretreatment according to the gravity;

[0108] 计算线性尺寸规整化后的各个自然笔画段长度的中心坐标,通过以下公式计算自然笔画段长度的中心坐标: [0108] NATURAL center coordinates of each stroke length of the linear regular size calculation, and calculating the natural length of the stroke center coordinates by the following equation:

[0109] [0109]

Figure CN104063705BD00111

[0110] X1为自然笔画段的起始横坐标值,χ1+1为自然笔画段中止横坐标值;Yi为自然笔画段的起点纵坐标值,γ1+1为自然笔画段的中止坐标的纵坐标值。 [0110] X1 is the starting abscissa value of natural stroke segment, χ1 + 1 is a natural suspension stroke segments abscissa value; Yi natural origin ordinate values ​​for stroke segment, γ1 + 1 is the suspension stroke segment of the natural coordinate longitudinal coordinate values.

[0111] 根据公式⑴和公式⑷得到笔迹数据的重心,笔迹数据的重心公式: [0111] handwriting data obtained according to the formula and the formula ⑷ ⑴ gravity, the center of gravity formula handwriting data:

[0112] [0112]

Figure CN104063705BD00112

[0113] 其中,上述重心公式中的各参数含义与公式⑴和⑷中相同。 [0113] wherein the same center of gravity of the above formula with the meaning of the parameters in equation ⑴ and ⑷.

[01Μ] 优选地,所述脱机特征包括投影脱机特征、或网格脱机特征、或扇形脱机特征、或轮廓脱机特征。 [01Μ] Preferably, the offline feature comprises a projection offline features, mesh or offline features, or fan-off features, characteristics or profile offline.

[0115] 优选地,当所述脱机特征为所述投影脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的脱机特征的步骤包括: [0115] Preferably, when the offline feature wherein said projection offline, the offline feature extraction after the pretreatment of handwriting data based on said center of gravity comprises the step of:

[0116] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上部分区域和下部分区域,从垂直方向分割为左部分区域和右部分区域。 [0116] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data divided from the horizontal direction to an upper portion and a lower region partial regions divided from a vertical direction to the left and right portions of the partial region area.

[0117] 以笔迹数据的重心为分割点的投影脱机特征如图5所示。 [0117] wherein the projection off the center of gravity of stroke data division point as shown in FIG. 其中,重心用实心的圆点表示,被分割后的区域用格子表示。 Wherein, the center of gravity is represented by a solid dot, the divided region indicated by a grid. 然后在分割后的上部分区域、下部分区域、左部分区域和右部分区域分别扫描各个笔画段向量的中心坐标在所述上部分区域、下部分区域、左部分区域和右部分区域出现的个数。 And then on a portion of the divided region, the lower partial region, a left region and a right portion, respectively, the partial region of each stroke segments vectors scanning center coordinates in a partial region of the upper, lower partial region, a left region and a right portion of a partial region appears number. 其中,分割为上部分区域和下部分区域中的笔迹数据按照从左到右的方式或者从右到左的方式进行扫描,分割为左部分区域和右部分区域中的笔迹数据按照从上到下的方式或者从下到上的方式进行扫描。 Wherein the region is divided into an upper portion and a lower portion of the stroke data in the area from left to right or right to left fashion way scanning, it is divided from top to bottom and left partial region of the stroke data in the right part area manner or scanning from the bottom-up manner.

[0118] 当所述脱机特征为所述网格脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的脱机特征的步骤包括: Step [0118] When the offline feature offline wherein said mesh, according to the center of gravity of the offline feature extraction after the pretreatment stroke data comprises:

[0119] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北,具体的八个方向如图6所示。 [0119] defined two-dimensional plane of the eight directions, east, west, south, north, southeast, northeast, southwest, northwest, eight specific direction as shown in Figure 6.

[0120] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上网格和下网格,从垂直方向分割为左网格和右网格。 [0120] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, the stroke data is divided into a grid on the pretreatment and the net horizontal direction cells divided from a vertical direction to the left and right grid mesh. 其中,上网格和下网格的格数一致,左网格和右网格的格数也一致;且上网格高度与下网格高度一致,左网格宽度与右网格宽度也一致。 Wherein the number of cells in the grid and the lower grid coincides number of squares left and right mesh grid is also the same; and the height of the grid on the same grid height, a left and right mesh width of the grid width is also consistent. 然后在分割后的上网格、下网格、左网格和右网格内分别扫描各个笔画段向量的中心坐标在所述上网格、下网格、左网格和右网格的八个方向上出现的个数。 Then the divided grid on, the grid, mesh left and right, respectively, scanning the center of each grid segment stroke vectors in eight directions of the coordinate in the grid, the grid, the left and right grid mesh the number appears on.

[0121] 当所述脱机特征为所述扇形脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的扇形脱机特征的步骤包括: [0121] wherein when the step of the offline feature when the fan off, according to the center of gravity of the offline feature extraction fan preprocessed stroke data comprises:

[0122] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0122] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0123] 以所述笔迹数据的重心为圆心,对所述预处理后的笔迹数据分割为多个扇形区域。 [0123] In the center of gravity as the center of the stroke data, handwriting data after the pretreatment is divided into a plurality of fan-shaped region. 例如:以重心为圆心,重心用实心的圆点表示(如图7中圆心处的实心圆点),笔迹数据被划分为16个扇形区域,如图7所示。 For example: as the center of gravity, the center of gravity is represented by solid dots (solid dots in FIG. 7 at the center), stroke data 16 is divided into fan-shaped area, as shown in FIG. 分别扫描被划分为16个扇形区域中的各个笔画段向量的中心坐标在八个方向上出现的个数。 Respectively, the center of each scan is divided into stroke segments vectors sector region 16 appearing in the number of coordinates in eight directions.

[0124] 当所述脱机特征为所述轮廓脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的轮廓脱机特征的步骤包括: Step [0124] When the off-line profile wherein the offline feature, the center of gravity based on the extracted outline feature offline after the pretreatment stroke data comprises:

[0125] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0125] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0126] 以所述笔迹数据的重心为结束点,分别扫描各个笔画段向量的中心坐标在八个方向出现的个数,其中,可以从东向西或者从西向东扫描,具体的扫描方式本申请不加以限制。 [0126] In the center of gravity of the stroke data is an end point, respectively, the scanning center coordinates of each stroke segments vectors occurring in the number of eight directions, which, from east to west or west to east scanned, scanning the specific embodiment of the present application is not limited thereto.

[0127] 需要说明的是,笔迹数据的重心为圆心,划分多个扇形区域,可以由本领域技术人员根据实际情况适当划分扇形区域,优选为16个扇形区域。 [0127] Incidentally, the center of gravity of stroke data, a plurality of fan-shaped area is divided, by those skilled in the art can be suitably divided fan-shaped area according to the actual situation, preferably 16 sector region.

[0128] 步骤306:依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 [0128] Step 306: according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition.

[0129] 依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 [0129] according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition.

[0130] 基于上述方法实施例的说明,本申请还提供了相应的一种笔迹特征提取装置的实施例,来实现上述方法实施例所述的内容。 [0130] Based on the above description of the embodiments of the method, the present application also provides a corresponding embodiment of handwriting feature extraction means, content is achieved according to the method described above in Example embodiment.

[0131] 参见图8,示出了本申请实施例四中的一种笔迹特征提取装置的结构框图,具体可以包括:获取模块,用于按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据。 [0131] Referring to Figure 8, there is shown a block diagram of apparatus according to the present application handwriting feature extracting one embodiment Fourth embodiment may specifically include: an obtaining module, for writing the time-series data in accordance with the handwriting stroke data acquisition and pre-processing , pretreatment of handwriting data obtained.

[0132] 分割模块,用于依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段, 获得多个笔画段向量。 [0132] The segmentation module configured to segment according to the time-series interval on the handwriting data for the pretreatment and the like, to obtain a plurality of stroke segments vectors.

[0133] 计算模块,用于获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标。 [0133] calculation means for obtaining the plurality of online feature stroke segments vectors, wherein said line includes an angle of the plurality of stroke segments vectors, and the center coordinates.

[0134] 提取模块,用于根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征。 [0134] extraction means for acquiring the center of gravity of the handwriting data based on the handwriting data of the pretreatment, the offline feature extraction stroke data after the pretreatment according to the center of gravity.

[0135] 处理模块,用于依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征。 [0135] The processing module configured according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition.

[0136] 优选地,所述获取模块包括:线性规整模块,用于将采集的笔迹数据按照书写的时间序列进行线性尺寸规整化后,获得各个自然笔画段长度。 [0136] Preferably, the obtaining module comprises: a linear structured module for the handwriting data collected after regular linear dimensions, to thereby obtain the respective stroke length naturally written in time series.

[0137] 长度获取模块,用于依据获得的所述各个自然笔画段长度,得到由所述各个自然笔画段组成的笔迹数据的长度。 [0137] length acquiring module, for each of the natural length of the stroke segments obtained according to obtain the length of the handwriting data by the respective segments of the natural stroke.

[0138] 优选地,所述多个笔画段向量的角度包括:各个笔画段向量与X轴正方向的角度、 各个笔画段向量与Y轴正方向的角度、以及,各个笔画段向量与其相邻的笔画段向量之间的角度。 [0138] Preferably, the plurality of angular stroke segments vectors comprising: an angle vector and the positive direction of the X-axis of each stroke segment, a vector angle of the Y-axis and the positive direction of each stroke segment, and a respective adjacent stroke segments vectors the angle between the stroke segments vectors.

[0139] 优选地,所述脱机特征包括投影脱机特征、或网格脱机特征、或扇形脱机特征、或轮廓脱机特征。 [0139] Preferably, the offline feature comprises a projection offline features, mesh or offline features, or fan-off features, characteristics or profile offline.

[0140] 优选地,当所述脱机特征为所述投影脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的脱机特征时: [0140] Preferably, when the offline wherein said projection offline feature, the feature extraction module extracts offline after the pretreatment according to the handwriting data in the center of gravity:

[0141] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上部分区域和下部分区域,从垂直方向分割为左部分区域和右部分区域,分别扫描各个笔画段向量的中心坐标在所述上部分区域、下部分区域、左部分区域和右部分区域出现的个数。 [0141] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data divided from the horizontal direction to an upper portion and a lower region partial regions divided from a vertical direction to the left area and the right partial region portion, respectively, the scanning center coordinates of each stroke segments vectors in the upper partial region, the number of lower partial region, a left region and a right portion of the partial region appears.

[0142] 当所述脱机特征为所述网格脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的脱机特征时: [0142] When the offline feature wherein said mesh offline, the offline feature extraction module extracts the stroke data after the pretreatment in accordance with the center of gravity:

[0143] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0143] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0144] 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上网格和下网格,从垂直方向分割为左网格和右网格,分别扫描各个笔画段向量的中心坐标在所述上网格、下网格、 左网格和右网格的八个方向上出现的个数。 [0144] In the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is divided horizontal direction and the vertical direction is divided, the stroke data is divided into a grid on the pretreatment and the net horizontal direction cells divided from a vertical direction to the left and right grid meshes respectively scanning center coordinates of each stroke segments vectors number appearing on the eight directions in the grid, grid lower, left, and right grid mesh .

[0145] 当所述脱机特征为所述扇形脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的扇形脱机特征时: [0145] When the feature is off when the fan off feature, the feature extraction module extracts the fan off after the pretreatment according to the handwriting data at the center of gravity:

[0146] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0146] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0147] 以所述笔迹数据的重心为圆心,对所述预处理后的笔迹数据分割为多个扇形区域,分别扫描各个笔画段向量的中心坐标在八个方向上出现的个数。 [0147] In the center of gravity as the center of the stroke data, handwriting data after the pretreatment is divided into a plurality of fan-shaped region, respectively, each of the scanning center coordinates of the stroke segments vectors appearing in the number of eight directions.

[0148] 当所述脱机特征为所述轮廓脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的轮廓脱机特征时: [0148] When the off-line profile wherein the offline feature, the feature extraction module extracts the outline of offline handwriting data after the pretreatment in accordance with the center of gravity:

[0149] 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; Eight directions [0149] defined two-dimensional plane, east, west, south, north, southeast, northeast, southwest, northwest;

[0150] 以所述笔迹数据的重心为结束点,分别扫描各个笔画段向量的中心坐标在八个方向出现的个数。 [0150] In the center of gravity of the stroke data is an end point, respectively, the scanning center coordinates of each stroke segments vectors occurring in the number of eight directions.

[0151] 综上所述,本申请实施例一种笔迹特征提取装置主要包括以下优点: [0151] In summary, the embodiment of a handwriting feature extraction means comprises a main embodiment of the present application the following advantages:

[0152] 首先,本申请依据时间序列对预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量的联机特征,所述联机特征包括多个笔画段向量的角度和中心坐标。 [0152] First, the present application for handwriting data pre-processed and the like based on the time sequence of the spacer segment, a plurality of stroke segments vectors online feature, the online feature comprises a plurality of angle and center coordinates of the stroke segments vectors. 通过计算多个笔画段向量的角度和中心坐标,从而使笔迹数据的特征提取覆盖了笔迹数据的局部特性和全局特性,避免了现有方法中仅考虑笔迹数据特征点的位置,从而造成笔迹数据特征提取不全面的问题。 By calculating the center coordinates and the angle of the plurality of stroke segments vectors, the stroke data so that the extracted feature local coverage characteristics and global characteristics of handwriting data, existing techniques only consider the position of the feature point handwriting data, the handwriting data resulting feature extraction incomplete question.

[0153] 其次,本申请通过对预处理后的笔迹数据获取笔迹数据的重心,并依据重心进行对称投影,然后提取相邻区域的笔迹数据的局部特征和全局特性,从而避免了等宽和等高方式进行提取笔迹数据特征时过于机械和形变适应性不佳的问题。 [0153] Next, the present application by pretreatment of handwriting data acquisition center of gravity of the handwriting data, based on projected center of gravity and symmetrically and extracts local features and global characteristics of the regions adjacent to the handwriting data, thus avoiding the width and the like when handwriting data feature extraction methods are too high mechanical strain and poor adaptability.

[0154] 再次,本申请通过对提取的联机特征和脱机特征的组合,得到了有效的笔迹数据特征,进而保证了后续分类器训练的可靠性,并显著提高了分类器的分类准确度,最终提高了联机手写的识别准确率。 [0154] Again, the present application features through a combination of online and offline feature extraction, has been effective feature stroke data, thereby ensuring the reliability of a subsequent classifier training, and significantly improves the classification accuracy of the classifier, and ultimately improve the accuracy of online handwriting recognition.

[0155] 对于装置实施例而言,由于其与方法实施例基本相似,所以描述的比较简单,相关之处参见方法实施例的部分说明即可。 [0155] For the apparatus of the embodiment, since the method of the embodiment which is substantially similar, the description of a relatively simple, some embodiments of the methods see relevant point can be described.

[0156] 本说明书中的各个实施例均采用递进的方式描述,每个实施例重点说明的都是与其他实施例的不同之处,各个实施例之间相同相似的部分互相参见即可。 [0156] In the present specification, various embodiments are described in a progressive way, differences from the embodiment and the other embodiments each of which emphasizes embodiment, the same portions similar between the various embodiments refer to each other.

[0157]以上对本申请所提供的一种笔迹特征提取的方法和装置,进行了详细介绍,本文中应用了具体个例对本申请的原理及实施方式进行了阐述,以上实施例的说明只是用于帮助理解本申请的方法及其核心思想;同时,对于本领域的一般技术人员,依据本申请的思想,在具体实施方式及应用范围上均会有改变之处,综上所述,本说明书内容不应理解为对本申请的限制。 [0157] The foregoing method and apparatus for handwriting features herein provide extracted, described in detail herein through specific examples of the principles and embodiments of the present application are set forth in description of the above embodiment except for help understanding the method and core ideas of the present application; Meanwhile, those of ordinary skill in the art based on the idea of ​​the present disclosure, may make modifications to the specific embodiments and applications of the embodiment, the summary, the content of the specification It should not be construed as limiting the present application.

Claims (8)

  1. 1. 一种笔迹特征提取的方法,其特征在于,包括: 按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据; 依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量; 获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标; 根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征; 依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征; 其中,所述脱机特征包括投影脱机特征; 当所述脱机特征为所述投影脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的脱机特征的步骤包括: 以所述笔迹数据的重心为分割点对所述预处 1. A method for handwriting feature extraction, which comprising: writing data in time series acquired in accordance with the stroke and pre-stroke data, stroke data obtained after pretreatment; according to the time sequence of the pre-processed interval segment handwriting data, etc., to obtain a plurality of stroke segments vectors; obtaining the plurality of online feature stroke segments vectors, wherein the line comprises a plurality of angles and the center coordinates of the stroke segments vectors; according to the pre- obtaining handwriting data processing center of gravity of the stroke data, according to the center of gravity of the offline feature extraction stroke data after the preprocessing; according to the features of online and offline features, numerical normalization, the said normalized value as a result of the characteristic data acquired handwriting; wherein said offline feature comprises a projection wherein the offline; offline when the offline feature wherein the projection, according to the said step offline feature stroke data after extracting said center of gravity of the pretreatment comprises: gravity center of the stroke data division point is at the pre- 后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上部分区域和下部分区域,从垂直方向分割为左部分区域和右部分区域,分别扫描各个笔画段向量的中心坐标在所述上部分区域、下部分区域、左部分区域和右部分区域出现的个数。 After the stroke data is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data area is divided into an upper portion and a lower partial regions divided from a vertical direction to the left and right partial regions from partial regions in the horizontal direction, respectively, the scanning Center coordinates of each stroke segments vectors in the upper partial region, the number of lower partial region, a left region and a right portion of the partial region appears.
  2. 2. 根据权利要求1所述的方法,其特征在于,所述按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据的步骤包括: 将采集的笔迹数据按照书写的时间序列进行线性尺寸规整化后,获得各个自然笔画段长度; 依据获得的所述各个自然笔画段长度,得到由所述各个自然笔画段组成的笔迹数据的长度。 2. The method according to claim 1, wherein said handwriting data written in accordance with the time-series data acquisition and pre-stroke, after the step of obtaining the pre-stroke data comprises: handwriting data collected in the writing after linear time series of regular size, to obtain the respective stroke length NATURAL; NATURAL stroke depending on the length of each of the segments obtained, the length of stroke data obtained by the respective segments of the natural stroke.
  3. 3. 根据权利要求1所述的方法,其特征在于,所述多个笔画段向量的角度包括:各个笔画段向量与X轴正方向的角度、各个笔画段向量与Y轴正方向的角度以及各个笔画段向量与其相邻的笔画段向量之间的角度。 3. The method according to claim 1, wherein the angle of said plurality of stroke segments vectors comprising: an angle vector and the positive direction of the X-axis of each stroke segments, each angular stroke segments vectors and the positive direction of Y-axis and the angle between the respective adjacent stroke segments vectors stroke segments vectors.
  4. 4. 根据权利要求1所述的方法,其特征在于,所述脱机特征可由投影脱机特征替换为网格脱机特征、或扇形脱机特征、或轮廓脱机特征; 当所述脱机特征为所述网格脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的脱机特征的步骤包括: 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上网格和下网格,从垂直方向分割为左网格和右网格,分别扫描各个笔画段向量的中心坐标在所述上网格、下网格、左网格和右网格的八个方向上出现的个数; 当所述脱机特征为所述扇形脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的扇形脱机特征的步骤包括: 定义二维平面 4. The method according to claim 1, wherein said offline feature may be replaced with a grid wherein the projection offline offline features, characteristics or fan off, or off-line profile wherein; when the offline wherein said mesh when the offline feature, according to the center of gravity of the offline feature extraction stroke data after the preprocessing step comprises: defining a two-dimensional plane of the eight directions, east, west, south, north , southeast, northeast, southwest, northwest; gravity center of the stroke data to the split point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data divided in the horizontal direction and for the grid on the grid, divided in the vertical direction to the left and right grid meshes, each individual stroke segments vectors scanning center coordinates in the grid on the lower grid, the left and right grid mesh eight the number appearing on direction; a step of the offline feature when the fan is off feature, according to the center of gravity of the offline feature extraction fan stroke data after the pretreatment comprises: a two-dimensional plane defined 八个方向,东、西、南、北、东南、东北、西南、西北; 以所述笔迹数据的重心为圆心,对所述预处理后的笔迹数据分割为多个扇形区域,分别扫描各个笔画段向量的中心坐标在八个方向上出现的个数; 当所述脱机特征为所述轮廓脱机特征时,所述依据所述重心提取所述预处理后的笔迹数据的轮廓脱机特征的步骤包括: 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; 以所述笔迹数据的重心为结束点,分别扫描各个笔画段向量的中心坐标在八个方向出现的个数。 Eight directions, east, west, south, north, southeast, northeast, southwest, northwest; gravity center of the stroke data, the handwriting data is divided into a plurality of fan-shaped regions pretreatment, respectively, each scanning stroke the number of vector coordinates of the central segment appearing in eight directions; wherein said offline when the offline profile feature, the center of gravity based on the extracted outline feature offline handwriting data preprocessed the steps include: two-dimensional plane defined eight directions, east, west, south, north, southeast, northeast, southwest, northwest; the center of gravity to the handwriting data for the end points, respectively scanning center coordinates of each stroke segments vectors in the number eight directions appear.
  5. 5. —种笔迹特征提取的装置,其特征在于,包括: 获取模块,用于按照笔迹数据书写的时间序列采集笔迹数据并进行预处理,获得预处理后的笔迹数据; 分割模块,用于依据所述时间序列对所述预处理后的笔迹数据进行等间隔分段,获得多个笔画段向量; 计算模块,用于获取所述多个笔画段向量的联机特征,所述联机特征包括所述多个笔画段向量的角度和中心坐标; 提取模块,用于根据所述预处理后的笔迹数据获取所述笔迹数据的重心,依据所述重心提取所述预处理后的笔迹数据的脱机特征; 处理模块,用于依据所述联机特征和脱机特征,进行数值归一化处理,将所述数值归一化处理的结果作为采集的所述笔迹数据的特征; 其中,所述脱机特征包括投影脱机特征,当所述脱机特征为所述投影脱机特征时,所述提取模块在依据所述重心提取所述 5 - Species handwriting feature extraction means, characterized by comprising: an obtaining module, for writing the time-series data in accordance with the handwriting stroke data acquisition and pre-processing to obtain handwriting data preprocessing; splitting module, according to the stroke time series data segment intervals for the pretreatment, to obtain a plurality of stroke segments vectors; calculating module, configured to obtain a plurality of stroke segments vectors online feature, the online feature comprises angle and center coordinates of a plurality of stroke segments vectors; extracting module, configured to obtain the center of gravity of the handwriting data based on the handwriting data of the pretreatment, the offline feature extraction stroke data after the pretreatment according to the center of gravity ; processing module, configured according to the features of online and offline features, numerical normalization, the normalized value as a result of the stroke characteristic data acquisition; wherein said offline feature offline feature comprises a projection, wherein when said projection off the offline feature extraction in the extraction module according to the center of gravity of the 处理后的笔迹数据的脱机特征时: 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上部分区域和下部分区域,从垂直方向分割为左部分区域和右部分区域,分别扫描各个笔画段向量的中心坐标在所述上部分区域、下部分区域、左部分区域和右部分区域出现的个数。 When handwriting data processed offline features: the center of gravity of said stroke data is a dividing point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, the handwriting data from the pretreatment level dividing direction of an upper portion and a lower region of the partial region is divided in the vertical direction to the left area and the right partial region portion, respectively, the scanning center coordinates of each stroke segments vectors in the upper partial region, the lower partial region, a left portion and a right partial region number of regions occur.
  6. 6. 根据权利要求5所述的装置,其特征在于,所述获取模块包括: 线性规整模块,用于将采集的笔迹数据按照书写的时间序列进行线性尺寸规整化后, 获得各个自然笔画段长度; 长度获取模块,用于依据获得的所述各个自然笔画段长度,得到由所述各个自然笔画段组成的笔迹数据的长度。 6. The apparatus as claimed in claim 5, wherein the obtaining module comprises: a linear structured module for the collected handwriting data be written in time series after the linear dimension regularization, to obtain the respective stroke length NATURAL ; length acquiring module, for each of the natural length of the stroke segments obtained according to obtain the length of the handwriting data by the respective segments of the natural stroke.
  7. 7. 根据权利要求5所述的装置,其特征在于,所述多个笔画段向量的角度包括:各个笔画段向量与X轴正方向的角度、各个笔画段向量与Y轴正方向的角度以及各个笔画段向量与其相邻的笔画段向量之间的角度。 7. The apparatus as claimed in claim 5, wherein the angle of said plurality of stroke segments vectors comprising: an angle vector and the positive direction of the X-axis of each stroke segments, each angular stroke segments vectors and the positive direction of Y-axis and the angle between the respective adjacent stroke segments vectors stroke segments vectors.
  8. 8. 根据权利要求5所述的装置,其特征在于,所述脱机特征可由投影脱机特征替换为网格脱机特征、或扇形脱机特征、或轮廓脱机特征; 当所述脱机特征为所述网格脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的脱机特征时: 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; 以所述笔迹数据的重心为分割点对所述预处理后的笔迹数据进行水平方向分割和垂直方向分割,将所述预处理后的笔迹数据从水平方向分割为上网格和下网格,从垂直方向分割为左网格和右网格,分别扫描各个笔画段向量的中心坐标在所述上网格、下网格、左网格和右网格的八个方向上出现的个数; 当所述脱机特征为所述扇形脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的扇形脱机特征时: 定义二维 8. The device according to claim 5, wherein said offline feature may be replaced with a grid wherein the projection offline offline features, characteristics or fan off, or off-line profile wherein; when the offline wherein said mesh off feature, when extracting the offline feature extraction according to the center of gravity of the stroke data in the preprocessing module: definition of a two-dimensional plane of the eight directions, east, west, south, north, southeast, northeast, southwest, northwest; gravity center of the stroke data to the split point of the stroke data after the pretreatment is performed is divided in the horizontal direction and the vertical direction is divided, and the pretreatment of handwriting data from the horizontal direction and is divided into a grid on the lower grid, dividing the vertical direction from left and right grid meshes, each individual stroke segments vectors scanning center coordinates in the grid on the lower grid, the left and right grid mesh eight the number appearing on directions; when the offline wherein said fan off feature, in the extraction module according to the center of gravity of the offline feature extraction fan stroke data after the pre-treatment: two defined dimension 面的八个方向,东、西、南、北、东南、东北、西南、西北; 以所述笔迹数据的重心为圆心,对所述预处理后的笔迹数据分割为多个扇形区域,分别扫描各个笔画段向量的中心坐标在八个方向上出现的个数; 当所述脱机特征为所述轮廓脱机特征时,所述提取模块在依据所述重心提取所述预处理后的笔迹数据的轮廓脱机特征时: 定义二维平面的八个方向,东、西、南、北、东南、东北、西南、西北; 以所述笔迹数据的重心为结束点,分别扫描各个笔画段向量的中心坐标在八个方向出现的个数。 Eight directions of the plane, the east, west, south, north, southeast, northeast, southwest, northwest; gravity center of the stroke data, handwriting data after the pretreatment is divided into a plurality of fan-shaped regions, each scan Center coordinates of each stroke segments vectors number appearing in eight directions; wherein said offline when the offline profile feature, the center of gravity extraction module extracts the handwriting data in the pretreatment according to the the contour offline features: two-dimensional plane defined eight directions, east, west, south, north, southeast, northeast, southwest, northwest; the center of gravity to the end point of stroke data, respectively scanning each stroke segments vectors the number of center coordinates appear in eight directions.
CN 201410247878 2014-06-05 2014-06-05 A method and apparatus for handwriting feature extraction CN104063705B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201410247878 CN104063705B (en) 2014-06-05 2014-06-05 A method and apparatus for handwriting feature extraction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201410247878 CN104063705B (en) 2014-06-05 2014-06-05 A method and apparatus for handwriting feature extraction

Publications (2)

Publication Number Publication Date
CN104063705A true CN104063705A (en) 2014-09-24
CN104063705B true CN104063705B (en) 2017-08-11

Family

ID=51551409

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201410247878 CN104063705B (en) 2014-06-05 2014-06-05 A method and apparatus for handwriting feature extraction

Country Status (1)

Country Link
CN (1) CN104063705B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1255685A (en) * 1998-11-27 2000-06-07 英业达集团(西安)电子技术有限公司 Handwritten character recognition system without strokes order
CN1658221A (en) * 2004-01-14 2005-08-24 国际商业机器公司 Method and apparatus for performing handwriting recognition by analysis of stroke start and end points

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000057349A1 (en) * 1999-03-24 2000-09-28 British Telecommunications Public Limited Company Handwriting recognition system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1255685A (en) * 1998-11-27 2000-06-07 英业达集团(西安)电子技术有限公司 Handwritten character recognition system without strokes order
CN1658221A (en) * 2004-01-14 2005-08-24 国际商业机器公司 Method and apparatus for performing handwriting recognition by analysis of stroke start and end points

Also Published As

Publication number Publication date Type
CN104063705A (en) 2014-09-24 application

Similar Documents

Publication Publication Date Title
Pechwitz et al. Baseline estimation for Arabic handwritten words
Caesar et al. Preprocessing and feature extraction for a handwriting recognition system
Parui et al. Online handwritten Bangla character recognition using HMM
CN101853071A (en) Gesture identification method and system based on visual sense
Connell et al. Recognition of unconstrained online Devanagari characters
Bansal et al. On how to describe shapes of Devanagari characters and use them for recognition
Mahmoud Recognition of writer-independent off-line handwritten Arabic (Indian) numerals using hidden Markov models
Sharma et al. Online handwritten Gurmukhi character recognition using elastic matching
CN101989326A (en) Human posture recognition method and device
CN102073870A (en) Method for recognizing Chinese character handwriting on touch screen
Thomas et al. Synthetic handwritten captchas
Daifallah et al. Recognition-based segmentation algorithm for on-line arabic handwriting
Elbaati et al. Arabic handwriting recognition using restored stroke chronology
Kosmala et al. On-line handwritten formula recognition using statistical methods
CN1664846A (en) On-line hand-written Chinese characters recognition method based on statistic structural features
Varga et al. Template-based synthetic handwriting generation for the training of recognition systems
Agnihotri Offline Handwritten Devanagari Script Recognition
Fink et al. Online Bangla word recognition using sub-stroke level features and hidden Markov models
CN101290659A (en) Hand-written recognition method based on assembled classifier
Yu et al. Writer identification using dynamic features
Li et al. Writer identification of chinese handwriting using grid microstructure feature
CN101561866A (en) Character recognition method based on SIFT feature and gray scale difference value histogram feature
CN103092329A (en) Lip reading technology based lip language input method
Kishore et al. Video audio interface for recognizing gestures of indian sign
Mohammad et al. Optical character recognition implementation using pattern matching

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
GR01