A Dataset of Online Handwritten Assamese Characters
|
|
- Reynold Brown
- 5 years ago
- Views:
Transcription
1 J Inf Process Syst, Vol.11, No.3, pp.325~341, September ISSN X (Prnt) ISSN X (Electronc) A Dataset of Onlne Handwrtten Assamese Characters Udayan Baruah* and Shyamanta M. Hazarka** Abstract Ths paper descrbes the Tezpur Unversty dataset of onlne handwrtten Assamese characters. The onlne data acquston process nvolves the capturng of data as the text s wrtten on a dgtzer wth an electronc pen. A sensor pcks up the pen-tp movements, as well as pen-up/pen-down swtchng. The dataset contans 8,235 solated onlne handwrtten Assamese characters. Prelmnary results on the classfcaton of onlne handwrtten Assamese characters usng the above dataset are presented n ths paper. The use of the support vector machne classfer and the classfcaton accuracy for three dfferent feature vectors are explored n our research. Keywords Assamese, Character Recognton, Dataset Collecton, Data Verfcaton, Onlne Handwrtng, Support Vector Machne 1. Introducton Onlne handwrtng recognton s a major research topc because of emergng technologes, such as PDAs and tablet PCs. Onlne recognton refers to the methods and technques dealng wth the automatc processng of a character as t s wrtten usng a dgtzer [1]. Development of any recognton system requres a standard dataset, whch s used for the tranng and testng of the system. The acquston and dstrbuton of standard datasets have ncreasngly ganed mportance. Some examples of datasets n the onlne doman are the Englsh dataset Pen-Based Recognton of Handwrtten Dgts [2], the French dataset IRONOFF [3], the Spansh dataset UJIpenchars [4], and the UNIPEN dataset [5]. All of whch contan onlne handwrtng data from varous alphabets (ncludng Latn and Chnese), sgnatures, and pen gestures. Research on onlne handwrtng recognton has also emerged for Indan scrpts. Examples of a few Indan scrpts where onlne handwrtng recognton actvtes are beng carred out nclude Taml [6], Telugu [7], Bengal [8], Devanagar [9], and Gurmukh [10]. Few Indan scrpts have also reported standard datasets of onlne handwrtten characters. But not all datasets of Indan scrpts are known to be publcly avalable. Some examples of few datasets of onlne handwrtngs n the context of Indan scrpts are, the HP Lab Taml Dataset [11], the Bangla Numeral Dataset [12] and the Devanagar Character Dataset [13]. Assamese s a major scrpt n the northeastern part of Inda. No onlne Ths s an Open Access artcle dstrbuted under the terms of the Creatve Commons Attrbuton Non-Commercal Lcense ( whch permts unrestrcted non-commercal use, dstrbuton, and reproducton n any medum, provded the orgnal work s properly cted. Manuscrpt receved July 08, 2013; frst revson December 18, 2013; second revson March 13, 2014; accepted Aprl 16, 2014; onlnefrst October 20, Correspondng Author: Udayan Baruah (udayan.baruah10@gmal.com) * Dept. of Informaton Technology, Skkm Manpal Insttute of Technology, Skkm , Inda (udayan.baruah10@gmal.com) ** Dept. of Computer Scence and Engneerng, Tezpur Unversty, Assam , Inda (shyamanta@eee.org) Copyrght 2015 KIPS
2 A Dataset of Onlne Handwrtten Assamese Characters handwrtten Assamese characters dataset s avalable n any standard onlne repostory of datasets. Therefore, the major objectve s to buld a dataset of onlne handwrtten Assamese characters. The characterstcs of Assamese characters (ncludng some dstnct propertes unque to t) are lsted below. a. Assamese wrtng has evolved from the ancent Indan scrpt, Brahm. Brahm had varous classes, and Assamese orgnated from the Gupta Brahm scrpt [14]. b. In Assamese there are 10 numerc characters and 52 basc alphabetc characters that consst of 11 vowels and 41 consonants. c. An Assamese character set s unque n that s has a large number of conjunct consonants (Juktakkhor) of about [14]. d. Certan vowels, consonants, and conjunct consonants have headlnes called matra n Assamese. e. There s no upper and lower case wrtng n Assamese. Ths makes Assamese characters dstnct from Englsh characters. In ths paper, we report on the development of a dataset of onlne handwrtten Assamese characters. The onlne handwrtten Assamese characters dataset reported n ths paper contans a total of 8,235 onlne handwrtten Assamese characters from 183 classes, whch consst of Assamese numerals, basc alphabetc characters, and conjunct consonants. We also present the prelmnary results on the classfcaton of onlne handwrtten Assamese characters. The paper s organzed as follows: Secton 2 descrbes the acquston of data, whch ncludes dscussons on an expermental set-up for data acquston (ncludng the data acquston program and data acquston protocol) and the nspecton of collected data for errors. Secton 3 provdes a dscusson on dfferent attrbutes of the collected data. The expermental results of character recognton on the samples of collected data are presented n Secton 4 and we present our fnal comments n Secton Data Acquston Data acquston s the collecton of onlne handwrtten samples from dfferent wrters. Ths nvolves the recordng of samples on some handwrtng nput devces lke PDAs, tablet PCs, and pen-tablets. These devces record the trajectory of the wrtng. 2.1 Expermental Setup for Data Acquston For the creaton of an onlne handwrtten Assamese characters dataset, onlne handwrtten samples of 121 conjunct consonants, along wth the 10 numerc characters and 52 basc alphabetc characters for Assamese [15], were collected. The total number of samples correspondng to each wrter s 183 (= 52 basc alphabetc characters + 10 numerals conjunct consonants). Thus there are a total of 8,235 samples n the dataset wrtten by 45 wrters (= ). Prnted Assamese Numerals, Basc Alphabetc Characters (Vowels and Consonants), and a few nstances of Conjunct Consonants (Juktakkhor) are shown n Table 1. Fgs. 1 3 show samples of onlne handwrtten Assamese characters (the mages of all the Assamese characters gven as reference shapes can also be downloaded along wth the dataset from The mages of the Assamese characters n prnted form are documented n the fle Data_Table.pdf n the dataset). The onlne Assamese handwrtng samples were collected on an -ball 8060U pen-tablet that was 326 J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
3 Udayan Baruah and Shyamanta M. Hazarka connected to a laptop and ts cordless dgtal stylus pen was used through a GUI. A screenshot of the GUI s shown n Fg. 4. The resoluton of the pen-tablet was 2540 LPI. The process of data acquston was guded by an expermental protocol. The expermental protocol descrbes the systematc steps for the collecton of data, whch mnmze varatons n terms of the place of wrtng of characters, wrtng envronment, and wrtng postures. The samples of the dataset were collected from students, research scholars, and faculty members of Tezpur Unversty. The wrters were 45 volunteers belongng to dfferent age groups rangng from years old wth the average age beng that of 31. Table 1. Prnted assamese characters Assamese numerals Assamese vowels a আ ঈ u ঊ ঋ e ঐ o ঔ Assamese consonants ক খ গ ঘ ঙ চ ছ জ ঝ ঞ ট ঠ ড ঢ ণ ত থ দ ধ ন প ফ ব ভ ম য ৰ ল ৱ শ ষ স হ k ড় ঢ় য় A few nstances of assamese conjunct consonants (Juktakkhor) k k k k k m p s s n t l h s n s b m n Data acquston program The GUI n the data acquston program shows a text nput box on the screen along wth some other controls. The sze of the text nput box s 4,392 4,868 ponts, where the coordnates are nteger numbers rangng from 0 to 4,392 for X and 0 to 4,868 for Y, respectvely. The value of X goes left-torght and that of Y goes downwards, assumng that the orgn (0, 0) les on the left-top corner of the text nput box. The acquston program records the handwrtng as a stream of (X, Y) coordnate ponts usng the approprate pen poston sensor along wth the pen-up/pen-down swtchng. The pressure level values are not recorded. A sngle clck on an onscreen Save button saves the current content of the text nput box. Ths also clears the current content of the text nput box so that the wrter can wrte the next character n the box. The unsaved content of the text nput box can be cleared by usng an onscreen Reset button. The data collecton system s ntalzed to the recordng mode by usng an onscreen Start button. The recorded nformaton of each onlne handwrtten character s stored n a text fle. Each fle s named based on the par (M, N). The text fle M.N.txt represents the character wth the ID M wrtten by the wrter wth the ID N. For nstance, the fle txt represents the character wth the ID 132 wrtten by the wrter wth ID 10. The dstrbuton of the dataset conssts of 45 folders (one for each wrter) and a Data_Table.pdf fle. Ths fle contans nformaton about the character ID (ID), character name (Label) and the prnted shape of the character (Char). Each folder contans 183 text fles correspondng to the 183 characters wrtten by a sngle wrter Expermental protocol for data acquston The onlne Assamese handwrtng samples were collected n a laboratory envronment. As wrters J Inf Process Syst, Vol.11, No.3, pp.325~341, September
4 A Dataset of Onlne Handwrtten Assamese Characters may not be at ease wth the wrtng nterface and the electronc pen, they were nstructed to practce by wrtng several characters n the text nput box before the actual recordng of characters started. After a wrter became famlar wth the onlne hand wrtng tools and the envronment, hs/her actual data recordng process took place accordng to the stages M0 to M4. Each wrter contrbuted wth hs/her handwrtng samples n a sngle, unnterrupted sesson. M0: The wrter s made to st n a relaxed poston wth the data collecton hardware setup n from of hm/her. M1: The lst of all 183 canddate characters s vewed by the wrter. M2: The wrter ntalzes the data collecton system. The correspondng wrter s ID and the ID of the frst character n the lst are entered through the GUI, whch s followed by a clck on the onscreen Start button. M3: (1) For each of the 183 characters (n the ascendng order of the seral number of the characters n the lst provded) startng at the frst character, the followng two consecutve steps are performed by the wrter: a: The character s wrtten n the text nput box of the GUI. b: The content of the text nput box s saved by usng the onscreen Save button. (2) If a mstake or unsatsfactory wrtng of any of the k (for k = 1, 2, 3,,183) characters s notced mmedately after the character s wrtten, but before t s saved, the followng two consecutve steps are performed by the wrter: a: The content of the text nput box s cleared by usng the onscreen Reset button. b: The steps of M3(1) are followed correspondng to the kth character. M4: All of the collected 183 samples are stored n the correspondng data folder. 2.2 Varablty n Wrtng Styles The shape of a character vares from wrter to wrter. Dfferent wrters wrte the same character dfferently. Smlarly, the wrters tend to wrte at dfferent locatons n the text nput box. Therefore, the characters of the same class are of varable shapes, varable szes, and a varable range of (X, Y) coordnate values. Some collected samples wrtten by dfferent wrters are shown n Fg. 1. The varablty of wrtng patterns of the same character s shown n Fg. 2. Fg. 1. Samples from the dataset of onlne handwrtten Assamese characters wrtten by dfferent wrters. In ths fgure we present samples from eght dfferent wrters. Fg. 2. Varablty of patterns of the same character wrtten by three dfferent wrters. 328 J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
5 Udayan Baruah and Shyamanta M. Hazarka Fg. 3. A screenshot of a part of the Vsual Data Inspecton Program. Fg. 4. Data acquston GUI. 2.3 Vsual Data Inspecton The vsual nspecton of data ams at detectng human as well as system errors. Examples of these types of errors are skppng some character, wrongly recordng some character, overwrtng a character, etc. A separate program was developed for vewng the samples collected mmedately after the completon of a data acquston sesson (refer to Fg. 3). Wth the help of ths program, t s possble to verfy data vsually n the presence of the wrter mmedately after hs/her sesson of data acquston and to overwrte the erroneous character fles by rewrtng those characters followng the expermental protocol. 3. Data Attrbutes Each character n the dataset contans attrbute nformaton. These attrbutes are the Name, Number of Strokes, and Sequence of Strokes assocated wth the character. J Inf Process Syst, Vol.11, No.3, pp.325~341, September
6 A Dataset of Onlne Handwrtten Assamese Characters 3.1 Character Name The frst lne of each sample s CHARACTER_NAME: Character (as shown for a sample character n Fg. 9 n Appendx A). The Character s the Name of any one of the 183 characters. 3.2 Number of Strokes n a Sample A stroke s a sequence of ponts from the pen-down to the pen-up events. The total number of strokes used to wrte a character s represented by the lne STROKE_COUNT: Number refer to Fg. 9 n Appendx A), where Number s the total count of the strokes n the character. 3.3 Sequence of Strokes Each stroke begns wth the PEN_DOWN nformaton and the PEN_UP nformaton s followed by the PEN_DOWN nformaton between the two consecutve strokes. The end of a sample s represented by the PEN_UP nformaton, whch s followed by the END_CHARACTER: Character nformaton. Each stroke conssts of a sequence of X and Y coordnates values, whch are gven n the frst and the second columns of the text fle, respectvely. Correspondng to each par of values of X and Y coordnates, the STYLUS_STATE and STROKE nformaton s gven n the thrd and the fourth columns, of the text fle, respectvely (refer to Fg. 9 n Appendx A). The STYLUS_STATE s ether 1 or 0. Correspondng to each recorded (X, Y) pont, the STYLUS_STATE s 1 and correspondng to the PEN_UP nformaton the STYLUS_STATE s 0. STYLUS_STATE s kept blank correspondng to each pece of PEN_DOWN nformaton. The STROKE nformaton represents the seral number of the consttuent stroke of a sample. 4. Character Recognton Character recognton experments were performed on the 10 numeral classes (a total of 45 10=450 numerals) and 52 classes of basc alphabetc characters (a total of 45 52=2340 basc alphabetc characters) avalable n the dataset. The archtecture for the classfcaton of characters s shown n Fg. 5. In order to conduct the classfcaton stage, we explored the support vector machne (SVM) [16] wth three dfferent kernels lnear, polynomal, and radal bass functon (RBF). Dataset of onlne handwrtten Assamese characters Preprocessng Sze Normalzaton Smoothenng Resamplng Preprocessed character Computaton of features Feature Set SVM classfer Classfcaton results Fg. 5. Archtecture for the classfcaton of characters. 330 J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
7 Udayan Baruah and Shyamanta M. Hazarka 4.1 Data Preprocessng Sze normalzaton Onlne handwrtten characters are sequences of two-dmensonal coordnate ponts (refer to Fg. 10 n Appendx A). Snce the characters are wrtten n dfferent szes at dfferent locatons of the text nput box, all of the characters have dfferent ranges of values for the X and Y coordnates. Therefore, the dfferent ranges of values for the coordnate ponts need to be normalzed to a unform range of values. Ths step s performed by normalzng the values of the X and Y coordnates of the characters to the range (0, 200). Fg. 6(a) shows the orgnal character and the character after sze normalzaton s shown n Fg. 6(b). (a) Fg. 6. (a) Orgnal character and (b) character after sze normalzaton. (b) Smoothng The smoothng of characters s requred to reduce the sharp changes or roughness (jtters) from the stroke. It removes jtters from the data whle preservng underlyng patterns. Here, the smoothng of the characters was performed usng a 3-pont movng average flter n order to remove jtters from the characters. Fg. 7(b) represents the smoothng result of the character shown n the Fg. 7(a). (a) Fg. 7. (a) Orgnal character and (b) character after smoothng. (b) Resamplng An onlne handwrtten character s a strng of coordnate ponts. Dfferent characters have a dfferent number of ponts along the trajectory from the start to the end. Therefore, dfferent characters need to be resampled to fxed number ponts. Here, all of the characters have been resampled to a fxed number of 100 ponts. Fg. 8(b) represents the resamplng of the character shown n the Fg. 8(a) to 100 ponts. (a) Fg. 8. (a) Orgnal character and (b) character after resamplng. (b) J Inf Process Syst, Vol.11, No.3, pp.325~341, September
8 A Dataset of Onlne Handwrtten Assamese Characters 4.2 Feature Sets We wll now present the classfcaton results, whch are based on the followng feature vectors that were computed from the characters taken from the dataset above. The feature set FS:2 s based on [17] and the feature set FS:3 s an enhancement of the feature set FS:2. The feature set FS:3 ncludes all the features of FS:2 along wth fve other features Feature set FS:1 X and Y coordnate values of the resampled characters. The values of the X and Y coordnates of the characters are dvded nto sxty-four zones or subntervals of the normalzed range of values (0, 200) of X and Y. The number of ponts n a character fallng nto each sub-nterval of (0, 200) s consdered as a feature of the character (zone feature). Coordnates of the start and end ponts of each character. Dstance between the start and the end ponts of the character. Number of strokes n the character. Mean of the X and Y coordnate values. Standard devaton of X and Y coordnates values. Headlne feature Feature set FS:2 Normalzed horzontal coordnates. Normalzed vertcal coordnates. Drecton angle wth reference to the horzontal axs (cosne of the angle). Drecton angle wth reference to the vertcal axs (sne of the angle). Curvature wth reference to the horzontal axs. Curvature wth reference to the vertcal axs. Poston of the stylus (up or down) Feature set FS:3 Normalzed horzontal coordnates. Normalzed vertcal coordnates. Drecton angle wth reference to the horzontal axs (cosne of the angle). Drecton angle wth reference to the vertcal axs (sne of the angle). Curvature wth reference to the horzontal axs. Curvature wth reference to the vertcal axs. Poston of the stylus (up or down). The number of ponts contaned n each of the sxty-four subntervals of the character (zone feature). Dstance between the start and the end ponts of the character. Number of strokes n the character. Mean of the X and Y coordnate values. Standard devaton of the X and Y coordnate values. 332 J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
9 Udayan Baruah and Shyamanta M. Hazarka 4.3 SVM Based Classfcaton Experments usng polynomal kernel and Gaussan radal bass functons, whch are the common choces for kernel functons n SVM, were performed for the classfcaton of both numerals and basc alphabetc characters. A k-fold cross valdaton procedure was used for the tranng and testng of the SVM classfer wth varous parametrc settngs. In k-fold cross-valdaton, the orgnal sample was randomly parttoned nto k equal sze subsamples. Of the k subsamples, a sngle subsample was retaned as the valdaton data for testng the model, and the remanng k 1 subsamples was used as tranng data. The cross-valdaton process was then repeated k tmes, n whch each of the k subsamples were used exactly once as the valdaton data. The k results from the folds were then combned to produce a sngle estmaton. All of the observatons were used for both tranng and valdaton, and each observaton was used for valdaton exactly one tme. 10-fold cross-valdaton was done n ths work Support vector machnes SVMs (Support Vector Machnes) are a useful technque for data classfcaton [16]. The goal of SVM s to produce a model (based on the tranng data), whch predcts the target values of the test data gven only the test data attrbutes. n Gven a tranng set of nstance-label pars x, y ), 1,......, l where x R and y } ( the SVMs requre that a soluton s found for the followng optmzaton problem: l { 1, 1, mn w, b, subject to 1 2 T y ( w ( x ) b) 1, 0 T w w C l 1 x are mapped nto a hgher dmensonal space by the functon. SVM Here, the tranng vectors fnds a lnear separatng hyper plane wth the maxmal margn n ths hgher dmensonal space. C 0 T s the penalty parameter of the error term. Furthermore, K ( x, x j ) ( x ) ( x j ) s called the kernel functon. The three types of kernels that are used the most are as follows: T Lnear: K ( x, x j ) x x j T d Polynomal: K ( x, x j ) ( x x j r), 0 2 Radal Bass Functon (RBF): K ( x, x ) exp( x x ), 0 j Here,, r and d are the kernel parameters. In ths experment, we have used the SVM wth the lnear, polynomal, and RBF kernels. j 4.4 Expermental Results The overall recognton rates acheved for the onlne handwrtten assamese basc alphabetc characters were 71.54% (usng lnear kernel), 75.39% (usng polynomal kernel), and 78.38% (usng RBF kernel) wth a 10 fold cross valdaton process based on the feature set FS:1 (refer to Table 2). A total of 2,340 characters were used as samples n the basc alphabetc character recognton experment. The kernel J Inf Process Syst, Vol.11, No.3, pp.325~341, September
10 A Dataset of Onlne Handwrtten Assamese Characters Table 2. Statstcs for the recognton rates Type of SVM kernel Across classes Total no. Correctly of classfed Average SD of the nstances nstances recognton recognton Average±SD rate (%) rate (%) Onlne handwrtten Assamese basc alphabetc characters Feature set FS:1 Lnear (C=1, E=1) ±15.77 Polynomal (C=3, E=4) ±14.39 RBF (C=9, Gamma=0.07) ±13.31 Feature set FS:2 Lnear (C=1, E=1) ±14.29 Polynomal (C=3, E=4) ±12.23 RBF (C=8, Gamma= ) ±11.50 Feature set FS:3 Lnear (C=1, E=1) ±13.45 Polynomal (C=3, E=4) ±12.24 RBF (C=8, Gamma= ) ±11.08 Onlne handwrtten Assamese numerals Feature set FS:1 Lnear (C=1, E=1) ±2.21 Polynomal (C=1, E=4) ±1.54 RBF(C=3, Gamma=0.05) ±1.16 Feature set FS:2 Lnear (C=1, E=1) ±5.37 Polynomal (C=1, E=4) ±2.21 RBF (C=8, Gamma= ) ±1.79 Feature set FS:3 Lnear (C=1, E=1) ±3.32 Polynomal (C=1, E=4) ±2.56 RBF (C=8, Gamma= ) ±1.79 Table 3. Recognton rates of onlne handwrtten Assamese basc alphabetc characters based on FS:1 for dfferent combnatons of the polynomal kernel parameters C and E C (complexty parameter) E (exponent) Recognton rate (%) J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
11 Udayan Baruah and Shyamanta M. Hazarka parameter settng C = 3 and E = 4, whch are assocated wth the polynomal kernel, was obtaned by varyng dfferent values of C and E wthn a range and gettng the recognton rate at each combnaton of parameters. Here C s the complexty (or penalty) parameter and E s the exponent of the polynomal kernel. The recognton rates of the basc alphabetc characters based on FS:1 for dfferent combnatons of the parameters C and E assocated wth the polynomal kernel are shown n Table 3. Smlarly, the kernel parameter settng C = 9 and gamma = 0.07 assocated wth RBF kernel was obtaned by varyng dfferent values of C and gamma wthn a range and gettng the recognton rate at each combnaton of parameters. Smlarly, based on FS:2 [17], the overall recognton rate acheved was 80.29% and based on FS:3, the overall recognton rate acheved was 81.15% (refer to Table 2). The ndvdual recognton rates (n percentage) of the basc alphabetc characters class are represented n Table 4. Table 4. Indvdual recognton rates of the onlne handwrtten Assamese basc alphabetc characters Sl. no. Class Feature set FS:1 RBF kernel wth C=9 & Gamma= 0.07 Recognton rate (%) Feature set FS:2 RBF kernel wth C=8 & Gamma= Feature set FS:3 RBF kernel wth C=8 & Gamma= Sl. no. Class Feature set FS:1 RBF kernel wth C=9 & Gamma= 0.07 Recognton rate (%) Feature set FS:2 RBF kernel wth C=8 & Gamma= Feature set FS:3 RBF kernel wth C=8 & Gamma= A TA AA THA E DA EE DHA U NA UU PA REE PHA AE BA OI BHA O MA OU AJA KA RA KHA LA GA WA GHA TXA NG MXA CA DXA CCA HA JA KHYA JHA AYA NIYA DRA MTA DHRA MTHA KTA MDA ANSR MDHA BXG MNA CBN The overall recognton rate acheved for the onlne handwrtten Assamese numerals was 99.11% (refer to Table 2), whch was obtaned by usng the polynomal kernel wth the kernel parameter settngs C = 1 and E = 4 and a 10 fold cross valdaton process based on FS: Assamese numerc J Inf Process Syst, Vol.11, No.3, pp.325~341, September
12 A Dataset of Onlne Handwrtten Assamese Characters characters were used as samples n the experment of numeral recognton. The parameter settng C = 1 and E = 4 was obtaned by varyng dfferent values of C and E wthn a range and gettng the recognton rate at each combnaton of parameters. Smlarly, based on the feature set FS:2 [17], the overall recognton rate acheved was 98.00% (refer to Table 2) and based on FS:3, the overall recognton acheved was 97.78% (refer to Table 2). The ndvdual recognton rates (n percentage) of the numeral classes are represented n Table 5. Table 5. Indvdual recognton rates of onlne handwrtten Assamese numerals Recognton rate (%) Sl. Class Feature set FS:1 Feature set FS:2 Feature set FS:3 No. Polynomal kernel wth Polynomal kernel wth RBF kernel wth C=1 & E=4 C=1 & E=4 C=8 & Gamma= SUNYA EK DUI TINI CARI PAC CAY XAT ATH NAA Shape Analyss of Smlar Characters The overall recognton rates acheved for the onlne handwrtten Assamese basc alphabetc characters s not very encouragng. Ths may be because the major strokes of several characters are almost smlar to each other. Accordngly, there are chances of these characters beng msclassfed, whch would result n a low recognton rate. Table 6 llustrates some of these characters and each row n Table 6 represents smlar characters from dfferent classes. Table 6. Smlar characters from dfferent classes Smlar characters GHA AJA BHA MDA MA THA CA MDHA MNA PA 336 J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
13 Udayan Baruah and Shyamanta M. Hazarka Improvement n the recognton rates (refer to Table 4) of some of these characters has been notced (from feature set FS:1 through feature set FS:3) when the drecton angle feature and curvature feature along wth zone feature are used. Identfyng the dscrmnatng features for an mproved classfcaton of these types of smlar characters s part of ongong research. 5. Fnal Comments In ths work, we have reported on the development of a dataset of onlne handwrtten Assamese characters. The samples were collected from a varety of wrters belongng to varous groups, n order to acheve a varety of wrtng patterns of characters. The samples of the dataset were not preprocessed. The recorded nformaton was drectly stored n raw format, whch provded a scope for applyng dfferent preprocessng technques to the dataset. The dataset of onlne handwrtten Assamese characters can be downloaded for free from the UCI Machne Learnng Repostory [18]. Ths reported dataset s the only publcly avalable dataset of onlne handwrtten Assamese characters. The dataset ams at provdng samples for research n onlne handwrtng recognton for Assamese scrpts. The prelmnary results on the recognton of numerals and basc alphabetc characters taken from the dataset are reported n ths work. From among the three SVM kernels used n ths experment on numerals, the polynomal kernel gves the best recognton rate of 99.11% based on the feature set FS:1. The recognton rate obtaned n the case of onlne handwrtten Assamese numerals s hgher than the recognton rate of 96.6% obtaned n an HMM based technque reported n [19]. From among the three SVM kernels used n ths experment for Assamese alphabetc characters, the RBF kernel gves the best recognton rate of 81.15% based on the feature set FS:3. No results are avalable n the lterature on the topc of recognton of onlne handwrtten Assamese alphabetc characters. Explorng a robust set of features for mprovng the recognton of characters and extendng the same for recognton of conjunct consonants (Juktakkhor) s part of ongong research. Acknowledgement The Onlne Handwrtten Assamese Characters Dataset s the result of onlne handwrtng samples contrbuted by students, research scholars, and faculty members of Tezpur Unversty. Ther support s gratefully acknowledged. References [1] R. Plamondon and S. Srhar, Onlne and offlne handwrtng recognton: a comprehensve survey, IEEE Transactons on Pattern Analyss and Machne Intellgence, vol. 22, no. 1, pp , [2] E. Alpaydn and Fevz. Almoglu, Pen-based recognton of handwrtten dgts dataset, 1998; edu/ml/datasets/pen-based+recognton+of+handwrtten+dgts. [3] C. Vvard-Gaurdn, P. M. Lallcan, S. Knerr, and P. Bnter, IRESTE On/Off (IRONOFF) dual handwrtng database, n Proceedng of the 5th Internatonal Conference on Document Analyss and Recognton, Bangalore, Inda, 1999, pp J Inf Process Syst, Vol.11, No.3, pp.325~341, September
14 A Dataset of Onlne Handwrtten Assamese Characters [4] D. Llorens, F. Prat, A. Marzal, J. M. Vlar, M. J. Castro, J. C. Amengual, S. Barrachna, A. Castellanos, S. España, J. A. Gómez, J. Gorbe, A. Gordo, V. Palazón, G. Pers, R. Ramos-Garjo, and F. Zamora, The UJIpenchars Database: a pen-based database of solated handwrtten characters, n Proceedng of the 6th Internatonal Conference on Language Resources and Evaluaton (LREC), Marrakech, Morocco, 2008, pp [5] I. Guyon, L. Schomaker, R. Plamondon, M. Lberman, and S. Janet, UNIPEN project of on-lne data exchange and recognzer benchmarks, n Proceedng of the 12th IAPR Internatonal Conference on Pattern Recognton, Jerusalem, Israel, 1994, pp [6] A. Bharath and S. Madhvanath, Hdden Markov model for onlne handwrtten Taml word recognton, n Proceedngs of the 9th Internatonal Conference on Document Analyss and Recognton, Curtba, Brazl, 2007, pp [7] L. Prasanth, J. Babu, R. Sharma, and P. Rao, Elastc matchng of onlne handwrtten Taml and Telegu scrpts usng local features, n Proceedngs of the 9th Internatonal Conference on Document Analyss and Recognton, Curtba, Brazl, 2007, pp [8] U. Bhattacharya, B. K. Gupta, and S. K. Paru, Drecton code based features for recognton of onlne handwrtten characters of Bangla, n Proceedngs of the 9th Internatonal Conference on Document Analyss and Recognton, Curtba, Brazl, 2007, pp [9] N. Josh, G. Sta, A. G. Ramakrshnan, V. Deepu, and S. Madhvanath, Machne Recognton of Onlne Handwrtten Devanagar Characters, n Proceedngs of the 8th Internatonal Conference on Document Analyss and Recognton, Seoul, Korea, 2005, pp [10] A. Sharma, R. Kumar, and R. K. Sharma, Onlne handwrtten Gurmukh character recognton usng elastc matchng, n Proceedngs of the Congress on Image and Sgnal Processng, Hanan, Chna, 2008, pp [11] A. S. Bhaskarabhatla and S. Madhvanath, Experences n collecton of handwrtng data for onlne handwrtng recognton n Indc scrpts, n Proceedngs of the 4th Internatonal Conference on Language Resources and Evaluaton, Lsbon, Portugal, [12] B. B. Chaudhur, A complete handwrtten numeral database of Bangla: a major Indc scrpt, n Proceedngs of the 10th Internatonal Workshop on Fronters n Handwrtng Recognton, France, [13] K. C. Santosh, C. Nattee, and Bart Lamroy, Relatve postonng of stroke based clusterng: a new approach to on-lne handwrtten Devanagar character recognton, Internatonal Journal of Image & Graphcs (IJIG), vol. 12, no. 2, [14] N. Sahara and K. M. Konwar, LutPad: a fully uncode compatble Assamese wrtng software, n, Proceedngs of the 2nd Workshop on Advances n Text Input Methods (WTIM 2), Mumba, Inda, 2012, pp [15] N. S. Bhabendra, Amar Akhar (Second Part). Guwahat: Assam Book Hve, [16] C. W. Hsu, C. C. Chang, and C. J. Ln, A practcal gude to support vector classfcaton, Department of Computer Scence and Informaton Engneerng, Natonal Tawan Unversty, Tape, Tawan, [17] A. R. Ahmed, C. V. Gaudn, M. Khald, and R. Yusof, Onlne handwrtng recognton usng support vector machne, n Proceedngs of the 2nd Internatonal Conference on Artfcal Intellgence n Engneerng and Technology, 2004, Sabah, Malaysa, pp [18] U. Baruah and S. M. Hazarka, Onlne handwrtten Assamese characters dataset, 2011; datasets/onlne+handwrtten+assamese+characters+dataset. [19] G. S. Reddy, P Sharma, S. R. M. Prasanna, C. Mahanta and L. N. Sharma, Combned onlne and offlne Assamese handwrtten numeral recognzer, n Proceedngs of the Natonal Conference on Communcatons (NCC2012), Kharagpur, 2012, pp J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
15 Udayan Baruah and Shyamanta M. Hazarka Appendx A An llustraton on representaton of handwrtten characters n the dataset s gven below. The text fle correspondng to ths data format s txt, whch represents the sample wth name TTT, Character ID 77 wrtten by the wrter wth Wrter ID 37. The.TXT fle representaton of the character TTT correspondng to Wrter ID 37 s reproduced n Fg. 9. A plot of the (X, Y) ponts of the correspondng character s shown n the Fg. 10. From the character lstng Data_Table.pdf, t can be verfed that the name of the character s TTT. CHARACTER_NAME: TTT STROKE_COUNT: 2 X Y STYLUS_STATE STROKE PEN_DOWN PEN_UP 0 PEN_DOWN J Inf Process Syst, Vol.11, No.3, pp.325~341, September
16 A Dataset of Onlne Handwrtten Assamese Characters PEN_UP 0 END_CHARACTER: TTT Fg. 9. The.TXT fle representaton of the character TTT correspondng to Wrter ID 37. Fg. 10. Plot of the character TTT correspondng to Wrter ID J Inf Process Syst, Vol.11, No.3, pp.325~341, September 2015
17 Udayan Baruah and Shyamanta M. Hazarka Udayan Baruah He s an Assocate Professor of Informaton Technology at Skkm Manpal Insttute of Technology, Skkm, Inda. He receved hs M.Sc. n Mathematcs from the Faculty of Mathematcal Scences, North Campus, Unversty of Delh, Inda and M.Tech. n Informaton Technology from the Department of Computer Scence and Engneerng, Tezpur Unversty, Assam, Inda. Presently, he s workng towards hs Ph.D. degree n the Department of Computer Scence and Engneerng, Tezpur Unversty. Hs feld of research s Onlne Handwrtng Recognton. Shyamanta M. Hazarka He s a Professor of Computer Scence and Engneerng at Tezpur Unversty, Assam, Inda. He completed hs B.E. from Assam Engneerng College, Guwahat, Inda, M.Tech. n Robotcs from IIT Kanpur, Inda and Ph.D. from Unversty of Leeds, England. He has been a Full Professor for Cogntve Systems and Neuro Informatcs at the Cogntve Systems Group, Unversty of Bremen, Germany, for wnter Hs work has focused prmarly n knowledge representaton and reasonng and rehabltaton robotcs wth an am towards development of ntellgent assstve systems. J Inf Process Syst, Vol.11, No.3, pp.325~341, September
Parallelism for Nested Loops with Non-uniform and Flow Dependences
Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr
More informationSupport Vector Machines
/9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.
More informationShape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram
Shape Representaton Robust to the Sketchng Order Usng Dstance Map and Drecton Hstogram Department of Computer Scence Yonse Unversty Kwon Yun CONTENTS Revew Topc Proposed Method System Overvew Sketch Normalzaton
More information12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification
Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero
More informationContent Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers
IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth
More informationPERFORMANCE COMPARISON OF ONLINE HANDWRITING RECOGNITION SYSTEM FOR ASSAMESE LANGUAGE BASED ON HMM AND SVM MODELLING
PERFORMANCE COMPARISON OF ONLINE HANDWRIING RECOGNIION SYSEM FOR ASSAMESE LANGUAGE BASED ON HMM AND SVM MODELLING Deepoy Das, Rtuparna Dev, SRM Prasanna, Subhankar Ghosh, Krshna Nak Department of EEE,
More informationCluster Analysis of Electrical Behavior
Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School
More informationSkew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach
Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research
More informationLearning the Kernel Parameters in Kernel Minimum Distance Classifier
Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department
More informationBOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET
1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School
More informationMachine Learning 9. week
Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below
More informationA Binarization Algorithm specialized on Document Images and Photos
A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a
More informationEdge Detection in Noisy Images Using the Support Vector Machines
Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona
More informationFace Recognition Based on SVM and 2DPCA
Vol. 4, o. 3, September, 2011 Face Recognton Based on SVM and 2DPCA Tha Hoang Le, Len Bu Faculty of Informaton Technology, HCMC Unversty of Scence Faculty of Informaton Scences and Engneerng, Unversty
More informationPRÉSENTATIONS DE PROJETS
PRÉSENTATIONS DE PROJETS Rex Onlne (V. Atanasu) What s Rex? Rex s an onlne browser for collectons of wrtten documents [1]. Asde ths core functon t has however many other applcatons that make t nterestng
More information2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements
Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.
More informationAn Optimal Algorithm for Prufer Codes *
J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,
More informationClassification / Regression Support Vector Machines
Classfcaton / Regresson Support Vector Machnes Jeff Howbert Introducton to Machne Learnng Wnter 04 Topcs SVM classfers for lnearly separable classes SVM classfers for non-lnearly separable classes SVM
More informationCOMPARISON OF SUPPORT VECTOR MACHINE AND NEURAL NETWORK IN CHARACTER LEVEL DISCRIMINANT TRAINING FOR ONLINE WORD RECOGNITION
COMPARISON OF SUPPORT VECTOR MACHINE AND NEURAL NETWORK IN CHARACTER LEVEL DISCRIMINANT TRAINING FOR ONLINE WORD RECOGNITION Abdul Rahm Ahmad 1 Chrstan Vard-Gaudn 3 Marzuk Khald 2 Emle Posson 3 1 Unverst
More informationTN348: Openlab Module - Colocalization
TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages
More informationProblem Definitions and Evaluation Criteria for Computational Expensive Optimization
Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty
More informationClassifier Selection Based on Data Complexity Measures *
Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.
More informationImprovement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration
Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,
More informationAn Empirical Comparative Study of Online Handwriting Chinese Character Recognition:Simplified v.s.traditional
2013 12th Internatonal Conference on Document Analyss and Recognton An Emprcal Comparatve Study of Onlne Handwrtng Chnese Recognton:Smplfed v.s.tradtonal Yan Gao, Lanwen Jn +, Wexn Yang School of Electronc
More informationOutline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1
4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:
More informationThe Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole
Appled Mathematcs, 04, 5, 37-3 Publshed Onlne May 04 n ScRes. http://www.scrp.org/journal/am http://dx.do.org/0.436/am.04.584 The Research of Ellpse Parameter Fttng Algorthm of Ultrasonc Imagng Loggng
More informationEmpirical Distributions of Parameter Estimates. in Binary Logistic Regression Using Bootstrap
Int. Journal of Math. Analyss, Vol. 8, 4, no. 5, 7-7 HIKARI Ltd, www.m-hkar.com http://dx.do.org/.988/jma.4.494 Emprcal Dstrbutons of Parameter Estmates n Bnary Logstc Regresson Usng Bootstrap Anwar Ftranto*
More informationType-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data
Malaysan Journal of Mathematcal Scences 11(S) Aprl : 35 46 (2017) Specal Issue: The 2nd Internatonal Conference and Workshop on Mathematcal Analyss (ICWOMA 2016) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES
More informationRelated-Mode Attacks on CTR Encryption Mode
Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory
More informationLecture 5: Multilayer Perceptrons
Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented
More informationSLAM Summer School 2006 Practical 2: SLAM using Monocular Vision
SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,
More informationUsing Fuzzy Logic to Enhance the Large Size Remote Sensing Images
Internatonal Journal of Informaton and Electroncs Engneerng Vol. 5 No. 6 November 015 Usng Fuzzy Logc to Enhance the Large Sze Remote Sensng Images Trung Nguyen Tu Huy Ngo Hoang and Thoa Vu Van Abstract
More informationResearch and Application of Fingerprint Recognition Based on MATLAB
Send Orders for Reprnts to reprnts@benthamscence.ae The Open Automaton and Control Systems Journal, 205, 7, 07-07 Open Access Research and Applcaton of Fngerprnt Recognton Based on MATLAB Nng Lu* Department
More informationAn Approach to Real-Time Recognition of Chinese Handwritten Sentences
An Approach to Real-Tme Recognton of Chnese Handwrtten Sentences Da-Han Wang, Cheng-Ln Lu Natonal Laboratory of Pattern Recognton, Insttute of Automaton of Chnese Academy of Scences, Bejng 100190, P.R.
More informationClassification of Face Images Based on Gender using Dimensionality Reduction Techniques and SVM
Classfcaton of Face Images Based on Gender usng Dmensonalty Reducton Technques and SVM Fahm Mannan 260 266 294 School of Computer Scence McGll Unversty Abstract Ths report presents gender classfcaton based
More informationFEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur
FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents
More informationMULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION
MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and
More informationSum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints
Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan
More informationFeature Reduction and Selection
Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components
More informationSmoothing Spline ANOVA for variable screening
Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory
More informationSkew Angle Detection of Bangla Script using Radon Transform
Sew Angle Detection of Bangla Script using Radon Transform S. M. Murtoza Habib, Nawsher Ahamed Noor and Mumit Khan Center for Research on Bangla Language Processing, BRAC University, Dhaa, Bangladesh.
More informationth International Conference on Document Analysis and Recognition
2013 12th Internatonal Conference on Document Analyss and Recognton Onlne Handwrtten Cursve Word Recognton Usng Segmentaton-free n Combnaton wth P2DBM-MQDF Blan Zhu 1, Art Shvram 2, Srrangaraj Setlur 2,
More informationLoad Balancing for Hex-Cell Interconnection Network
Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,
More informationDetermining the Optimal Bandwidth Based on Multi-criterion Fusion
Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn
More informationAccumulated-Recognition-Rate Normalization for Combining Multiple On/Off-Line Japanese Character Classifiers Tested on a Large Database
4 th Internatonal Workshop on Multple Classfer Systems (MCS23) Guldford, UK Accumulated-Recognton-Rate Normalzaton for Combnng Multple On/Off-Lne Japanese Character Classfers Tested on a Large Database
More informationAn Image Fusion Approach Based on Segmentation Region
Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua
More informationQuick error verification of portable coordinate measuring arm
Quck error verfcaton of portable coordnate measurng arm J.F. Ouang, W.L. Lu, X.H. Qu State Ke Laborator of Precson Measurng Technolog and Instruments, Tanjn Unverst, Tanjn 7, Chna Tel.: + 86 [] 7-8-99
More informationSegmentation Based Online Word Recognition: A Conditional Random Field Driven Beam Search Strategy
2013 12th Internatonal Conference on Document Analyss and Recognton Segmentaton Based Onlne Word Recognton: A Condtonal Random Feld Drven Beam Search Strategy Art Shvram 1, Blan Zhu 2, Srrangaraj Setlur
More informationA Robust Authentication System Handwritten Documents using Local Features for Writer Identification
Regular Paper Journal of Computng Scence and Engneerng, Vol. 8, No., March 204, pp. -6 A Robust Authentcaton System Handwrtten Documents usng Local Features for Wrter Identfcaton Parves Kamal* and Fasal
More informationDetection of an Object by using Principal Component Analysis
Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,
More information3D vector computer graphics
3D vector computer graphcs Paolo Varagnolo: freelance engneer Padova Aprl 2016 Prvate Practce ----------------------------------- 1. Introducton Vector 3D model representaton n computer graphcs requres
More informationSupport Vector Machines
Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned
More informationThe Research of Support Vector Machine in Agricultural Data Classification
The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou
More informationIncremental Learning with Support Vector Machines and Fuzzy Set Theory
The 25th Workshop on Combnatoral Mathematcs and Computaton Theory Incremental Learnng wth Support Vector Machnes and Fuzzy Set Theory Yu-Mng Chuang 1 and Cha-Hwa Ln 2* 1 Department of Computer Scence and
More informationUser Authentication Based On Behavioral Mouse Dynamics Biometrics
User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA
More informationA New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1
A New Feature of Unformty of Image Texture Drectons Concdng wth the Human Eyes Percepton Xng-Jan He, De-Shuang Huang, Yue Zhang, Tat-Mng Lo 2, and Mchael R. Lyu 3 Intellgent Computng Lab, Insttute of Intellgent
More informationTsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance
Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for
More informationDescription of NTU Approach to NTCIR3 Multilingual Information Retrieval
Proceedngs of the Thrd NTCIR Workshop Descrpton of NTU Approach to NTCIR3 Multlngual Informaton Retreval Wen-Cheng Ln and Hsn-Hs Chen Department of Computer Scence and Informaton Engneerng Natonal Tawan
More informationy and the total sum of
Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton
More informationVRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) , Fax: (370-5) ,
VRT012 User s gude V0.1 Thank you for purchasng our product. We hope ths user-frendly devce wll be helpful n realsng your deas and brngng comfort to your lfe. Please take few mnutes to read ths manual
More informationCMPS 10 Introduction to Computer Science Lecture Notes
CPS 0 Introducton to Computer Scence Lecture Notes Chapter : Algorthm Desgn How should we present algorthms? Natural languages lke Englsh, Spansh, or French whch are rch n nterpretaton and meanng are not
More informationThe motion simulation of three-dof parallel manipulator based on VBAI and MATLAB Zhuo Zhen, Chaoying Liu* and Xueling Song
Internatonal Conference on Automaton, Mechancal Control and Computatonal Engneerng (AMCCE 25) he moton smulaton of three-dof parallel manpulator based on VBAI and MALAB Zhuo Zhen, Chaoyng Lu* and Xuelng
More informationIncremental MQDF Learning for Writer Adaptive Handwriting Recognition 1
200 2th Internatonal Conference on Fronters n Handwrtng Recognton Incremental MQDF Learnng for Wrter Adaptve Handwrtng Recognton Ka Dng, Lanwen Jn * School of Electronc and Informaton Engneerng, South
More informationCollaboratively Regularized Nearest Points for Set Based Recognition
Academc Center for Computng and Meda Studes, Kyoto Unversty Collaboratvely Regularzed Nearest Ponts for Set Based Recognton Yang Wu, Mchhko Mnoh, Masayuk Mukunok Kyoto Unversty 9/1/013 BMVC 013 @ Brstol,
More informationVol. 5, No. 3 March 2014 ISSN Journal of Emerging Trends in Computing and Information Sciences CIS Journal. All rights reserved.
Journal of Emergng Trends n Computng and Informaton Scences 009-03 CIS Journal. All rghts reserved. http://www.csjournal.org Unhealthy Detecton n Lvestock Texture Images usng Subsampled Contourlet Transform
More informationPositive Semi-definite Programming Localization in Wireless Sensor Networks
Postve Sem-defnte Programmng Localzaton n Wreless Sensor etworks Shengdong Xe 1,, Jn Wang, Aqun Hu 1, Yunl Gu, Jang Xu, 1 School of Informaton Scence and Engneerng, Southeast Unversty, 10096, anjng Computer
More informationCorner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity
Journal of Sgnal and Informaton Processng, 013, 4, 114-119 do:10.436/jsp.013.43b00 Publshed Onlne August 013 (http://www.scrp.org/journal/jsp) Corner-Based Image Algnment usng Pyramd Structure wth Gradent
More informationSignature and Lexicon Pruning Techniques
Sgnature and Lexcon Prunng Technques Srnvas Palla, Hansheng Le, Venu Govndaraju Centre for Unfed Bometrcs and Sensors Unversty at Buffalo {spalla2, hle, govnd}@cedar.buffalo.edu Abstract Handwrtten word
More informationHuman Face Recognition Using Generalized. Kernel Fisher Discriminant
Human Face Recognton Usng Generalzed Kernel Fsher Dscrmnant ng-yu Sun,2 De-Shuang Huang Ln Guo. Insttute of Intellgent Machnes, Chnese Academy of Scences, P.O.ox 30, Hefe, Anhu, Chna. 2. Department of
More informationMachine Learning: Algorithms and Applications
14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of
More informationBrushlet Features for Texture Image Retrieval
DICTA00: Dgtal Image Computng Technques and Applcatons, 1 January 00, Melbourne, Australa 1 Brushlet Features for Texture Image Retreval Chbao Chen and Kap Luk Chan Informaton System Research Lab, School
More informationAudio Content Classification Method Research Based on Two-step Strategy
(IJACSA) Internatonal Journal of Advanced Computer Scence and Applcatons, Audo Content Classfcaton Method Research Based on Two-step Strategy Sume Lang Department of Computer Scence and Technology Chongqng
More informationA Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines
A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría
More informationAn Accurate Evaluation of Integrals in Convex and Non convex Polygonal Domain by Twelve Node Quadrilateral Finite Element Method
Internatonal Journal of Computatonal and Appled Mathematcs. ISSN 89-4966 Volume, Number (07), pp. 33-4 Research Inda Publcatons http://www.rpublcaton.com An Accurate Evaluaton of Integrals n Convex and
More informationNovel Pattern-based Fingerprint Recognition Technique Using 2D Wavelet Decomposition
Mathematcal Methods for Informaton Scence and Economcs Novel Pattern-based Fngerprnt Recognton Technque Usng D Wavelet Decomposton TUDOR BARBU Insttute of Computer Scence of the Romanan Academy T. Codrescu,,
More informationCS246: Mining Massive Datasets Jure Leskovec, Stanford University
CS46: Mnng Massve Datasets Jure Leskovec, Stanford Unversty http://cs46.stanford.edu /19/013 Jure Leskovec, Stanford CS46: Mnng Massve Datasets, http://cs46.stanford.edu Perceptron: y = sgn( x Ho to fnd
More informationOptimizing Document Scoring for Query Retrieval
Optmzng Document Scorng for Query Retreval Brent Ellwen baellwe@cs.stanford.edu Abstract The goal of ths project was to automate the process of tunng a document query engne. Specfcally, I used machne learnng
More informationAssignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.
Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton
More informationThe Codesign Challenge
ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.
More informationSteps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices
Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between
More informationA Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems
A Unfed Framework for Semantcs and Feature Based Relevance Feedback n Image Retreval Systems Ye Lu *, Chunhu Hu 2, Xngquan Zhu 3*, HongJang Zhang 2, Qang Yang * School of Computng Scence Smon Fraser Unversty
More informationThe Shortest Path of Touring Lines given in the Plane
Send Orders for Reprnts to reprnts@benthamscence.ae 262 The Open Cybernetcs & Systemcs Journal, 2015, 9, 262-267 The Shortest Path of Tourng Lnes gven n the Plane Open Access Ljuan Wang 1,2, Dandan He
More informationFace Recognition University at Buffalo CSE666 Lecture Slides Resources:
Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural
More informationNews. Recap: While Loop Example. Reading. Recap: Do Loop Example. Recap: For Loop Example
Unversty of Brtsh Columba CPSC, Intro to Computaton Jan-Apr Tamara Munzner News Assgnment correctons to ASCIIArtste.java posted defntely read WebCT bboards Arrays Lecture, Tue Feb based on sldes by Kurt
More informationA Fast Visual Tracking Algorithm Based on Circle Pixels Matching
A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng
More informationNAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics
Introducton G10 NAG Fortran Lbrary Chapter Introducton G10 Smoothng n Statstcs Contents 1 Scope of the Chapter... 2 2 Background to the Problems... 2 2.1 Smoothng Methods... 2 2.2 Smoothng Splnes and Regresson
More informationA high precision collaborative vision measurement of gear chamfering profile
Internatonal Conference on Advances n Mechancal Engneerng and Industral Informatcs (AMEII 05) A hgh precson collaboratve vson measurement of gear chamferng profle Conglng Zhou, a, Zengpu Xu, b, Chunmng
More informationAvailable online at Available online at Advanced in Control Engineering and Information Science
Avalable onlne at wwwscencedrectcom Avalable onlne at wwwscencedrectcom Proceda Proceda Engneerng Engneerng 00 (2011) 15000 000 (2011) 1642 1646 Proceda Engneerng wwwelsevercom/locate/proceda Advanced
More informationA Fast Content-Based Multimedia Retrieval Technique Using Compressed Data
A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,
More informationAnalysis on the Workspace of Six-degrees-of-freedom Industrial Robot Based on AutoCAD
Analyss on the Workspace of Sx-degrees-of-freedom Industral Robot Based on AutoCAD Jn-quan L 1, Ru Zhang 1,a, Fang Cu 1, Q Guan 1 and Yang Zhang 1 1 School of Automaton, Bejng Unversty of Posts and Telecommuncatons,
More informationUB at GeoCLEF Department of Geography Abstract
UB at GeoCLEF 2006 Mguel E. Ruz (1), Stuart Shapro (2), June Abbas (1), Slva B. Southwck (1) and Davd Mark (3) State Unversty of New York at Buffalo (1) Department of Lbrary and Informaton Studes (2) Department
More informationD. Barbuzzi* and G. Pirlo
Int. J. Sgnal and Imagng Systems Engneerng, Vol. 7, No. 4, 2014 245 bout retranng rule n mult-expert ntellgent system for sem-supervsed learnng usng SVM classfers D. Barbuzz* and G. Prlo Department of
More informationHistogram of Template for Pedestrian Detection
PAPER IEICE TRANS. FUNDAMENTALS/COMMUN./ELECTRON./INF. & SYST., VOL. E85-A/B/C/D, No. xx JANUARY 20xx Hstogram of Template for Pedestran Detecton Shaopeng Tang, Non Member, Satosh Goto Fellow Summary In
More informationA PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION
1 THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Seres A, OF THE ROMANIAN ACADEMY Volume 4, Number 2/2003, pp.000-000 A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION Tudor BARBU Insttute
More informationX- Chart Using ANOM Approach
ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are
More informationEnhanced Face Detection Technique Based on Color Correction Approach and SMQT Features
Journal of Software Engneerng and Applcatons, 2013, 6, 519-525 http://dx.do.org/10.4236/jsea.2013.610062 Publshed Onlne October 2013 (http://www.scrp.org/journal/jsea) 519 Enhanced Face Detecton Technque
More informationMachine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)
Machne Learnng Support Vector Machnes (contans materal adapted from talks by Constantn F. Alfers & Ioanns Tsamardnos, and Martn Law) Bryan Pardo, Machne Learnng: EECS 349 Fall 2014 Support Vector Machnes
More informationA MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS
Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung
More informationClassifying Acoustic Transient Signals Using Artificial Intelligence
Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)
More informationQuality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation
Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on
More information