Online Handwritten Cursive Word Recognition Using Segmentation-free and Segmentation-based Methods

Size: px
Start display at page:

Download "Online Handwritten Cursive Word Recognition Using Segmentation-free and Segmentation-based Methods"

Transcription

1 3th Asan Conference on Pattern Recognton (ACPR2015), November 3-6, 2015, Kuala Lumpur, Malasa. Onlne Handwrtten Cursve Word Recognton Usng Segmentaton-free and Segmentaton-based Methods Blan Zhua, Art Shvramb, Venu Govndarajub and Masak Nakagawaa a Department of Computer and Informaton Scences, Toko Unverst Agrculture and Technolog, Toko, Japan b Center for Unfed Bometrcs and Sensors, Unverst at Buffalo, Buffalo, US {ashvram, govnd}@buffalo.edu {zhublan, nakagawa}@cc.tuat.ac.jp Abstract Ths paper descrbes a comparson between onlne handwrtten cursve word recognton usng segmentaton-free method and that usng segmentaton-based method. To search the optmal segmentaton and recognton path as the recognton result, we attempt two methods: segmentaton-free and segmentaton-based, where we expand the search space usng a character-snchronous beam search strateg. The probable search paths are evaluated b ntegratng character recognton scores wth geometrc characterstcs of the character patterns n a Condtonal Random Feld (CRF) model. Our methods restrct the search paths from the tre lexcon of words and precedng paths durng path search. We show ths comparson on a publcl avalable dataset (IAM-OnDB). 1. Introducton The development of pen-based or touch-based nput devces such as tablets and smart phones has led to a push towards more flud nteractons wth these electroncs. Realzng onlne handwrtten character recognton wth hgh performance s vtal, especall for applcatons such as the natural nput of text on smart phones, to provde a satsfactor user experence. Wthout character recognton cues, characters cannot be segmented unambguousl due to the fact that segmentaton ponts between characters are not obvous. A feasble wa to overcome the ambgut of segmentaton s called ntegrated segmentaton and recognton, whch s classfed nto segmentaton-free and segmentaton-based methods. The segmentaton-based method attempts to splt cursve words nto character patterns at ther true boundares and label the splt character patterns. A word pattern s over-segmented nto prmtve segments such that each segment comprses of a sngle character or part of a character. The segments are combned to generate canddate character patterns (formng a canddate lattce), whch are evaluated usng character recognton, ncorporatng geometrc and lngustc contexts. On the other hand, the segmentaton-free method avods problems assocated wth segmentaton, and utlzes the abltes of one-dmensonal structural models such as HMMs or MRFs for concatenatng character models to construct word models based on the provded lexcon of words durng recognton, and erases and selects segmentaton ponts usng character models. Onlne handwrtten word recognton usng a recurrent neural network [1] performs word recognton b contnuousl movng the nput wndow of the network across the frame sequence representaton of a word, thus generatng actvaton traces at the output of the network. These output traces are subsequentl examned to determne the ASCII strng(s) best representng the word mage. It has got a lot of attenton and dscusson whch s better of the segmentaton-free and segmentaton-based methods for handwrtten strng recognton. In ths paper, we attempt the two methods and make a comparson between them for onlne handwrtten cursve word recognton. We tr the two methods (segmentaton-free and segmentaton-based) to search the optmal path as the recognton result. We expand the search space usng a character-snchronous beam search strateg and the probable search paths are evaluated b a path evaluaton crteron n a CRF model. To evaluate character patterns, we combne a MRF model [2] wth a P2DBMN-MQDF (pseudo 2D b-moment normalzaton - modfed quadratc dscrmnant functon) recognzer [3]. The rest of ths paper s organzed as follows: Secton 2 begns wth the descrpton of the preprocessng steps. Secton 3 descrbes the CRFs for our word recognton, Secton 4 presents the two search methods (segmentaton-free and segmentaton-based), Secton 5 presents the expermental results, and Secton 6 makes our conclusons. 2. Preprocessng Gven a word, we compute baselnes usng lnear regresson lnes that approxmate the local mnma or the local maxma. Then, we normalze the dfferent slants of 225

2 Fgure 1: Extractng feature ponts and neghborhood graphs. words, b shearng ever word accordng to ts slant angle. After transformng ever word to a gven corpus heght (the dstance between the baselne and the corpus lne), we re-sample the trajector ponts to ensure that pen-tp coordnates are equdstant usng lnear nterpolaton. Then, smoothng s accomplshed b applng a Gaussan flter to each coordnate pont n the sequence. Crtcal ponts such as start and end ponts of strokes and ponts of hgh curvature are detected and ncluded n the re-sampled/smoothed trajector output. We also remove delaed strokes (e.g., crosses on t s, dots on s etc.). Fnall, we extract feature ponts usng the method developed b Ramner [2]. The start and end ponts of ever stroke are pcked up as feature ponts. Then, the most dstant pont from the straght lne between adjacent feature ponts s selected as a feature pont f the dstance to the straght lne s greater than a threshold value. Ths selecton s done recursvel untl no more feature ponts are selected. 3. Word recognton CRFs 3.1. CRFs Shvram et al. [4] have proposed a CRF model for onlne handwrtten word recognton, where the CRF model was constructed from over-segmented ponts. Here we appl a smlar model, but formulate t based on feature ponts to appl the segmentaton-free method. Consder an nput word pattern X whch has a sequence of feature ponts F={f 1, f 2, f 3,, f g } as shown n Fg. 1. There are man probable segmentatons {S 1, S 2, S 3,, S m }, and each S ={s 1, s 2, s 3,,s n_} where s j denotes a canddate character pattern between feature ponts p k and p l (k, l = 1~g ). For each S we construct a neghborhood graph g as shown n Fg. 1, where each node denotes a canddate character pattern and each lnk represents the relatonshp between neghborng canddate character patterns. Let (s j-1,s j) denote a par of neghborng canddate characters s j-1 and s j. s j and (s j-1,s j) correspond to unar (sngle) and bnar (par-wse) clques, respectvel. Assumng that Y refers to the label sequence of S, P(Y S,X) can be approxmated b a real-valued energ functon E(X,S,Y,Λ) wth clque set N and parameters Λ: exp( E( X, S, Y, Λ)) P( Y S, X ) = Z( S, X ) Z( S, X ) = ( Y ') exp( E( X, S, Y', Λ)) Z(S,X), the normalzaton constant ma be gnored f we do not requre strct probablt values. Then, the problem of fndng the best label Y, that nvolves maxmzng (1) P(Y S,X), becomes equvalent to mnmzng the total graph energ: Y = arg max P(Y S,X) = arg mn E(X,S,Y,Λ) (2) To maxmze effcenc we utlze onl unar and bnar clques. Thus, the total energ functon s defned as k E( X, S, Y, Λ) = λk f( s j, s j )( j 1, j ) 1 ( s j =, s j ) S k 1 1 (3) k where f (, ( s, ) 1 ) 1 s j are feature functons on a bnar j j j clque (s j-1,s j), Λ={λ k } are weghtng parameters, j-1 and j denote the labels of s j-1 and s j, respectvel. In eq. (3) we consder onl bnar clques (s j-1,s j) snce the subsume the feature functons of unar clques s j. The total energ functon s used to measure the plausblt of Y, and the smaller E(X,S,Y,Λ) s, the larger wll P(Y S,X) be. For all segmentatons {S 1, S 2, S 3,, S m }, we select the best path. Denotng the number of feature ponts on S b N(S, X), we use the followng path evaluaton crteron to evaluate all segmentaton and recognton paths. (4) EC E ( X, S, Y, N, Λ ) ( S, Y, X ) = N ( S, X ) N(S, X) s the same for all segmentaton hpotheses on the total path length, and t would dffer whle evaluatng partal paths n a beam search. Therefore, t s necessar to normalze the path scores usng N(S, X) when evaluatng partal paths n a beam search. When usng a Vterb search such as dnamc programmng (DP) t would make the model analogous to a sem-crf framework as n [5] Feature functons The feature functons are unar f defned on sngle clques, and bnar on par-wse clques. On the other hand, the feature functons are class-relevant f dependent on character class (or class-par), otherwse are class-rrelevant. So, there are totall four tpes of feature functons: unar class-relevant, unar class-rrelevant, bnar class-relevant and bnar class-rrelevant. There are two output scores of character recognzers on canddate character patterns (two unar class-relevant feature functons): a P2DBMN-MQDF classfer on drecton hstogram features [3] and a MRF classfer [2]. The P2DBMN-MQDF classfer evaluates shapes of character patterns whle the MRF classfer evaluates structures of character patterns. The feature ponts extracted (secton 2) are normalzed so that the mnmum of X coordnate s 0 whle Y coordnates are unchanged, and, are used to construct one-dmensonal structural character MRFs to perform recognton. On the P2DBMN-MQDF classfer, feature vector dmensonalt s reduced usng Fsher Lnear Dscrmnant Analss (FLDA). There are few smbol classes n Englsh (78 for our dataset) and whle usng FLDA the reduced dmensonalt must be less than the number of smbol classes. Usng too few a feature dmenson results n low recognton accurac. To resolve K 226

3 ths, we create sub-classes for each smbol categor b applng k-means and use these sub-class categores nstead of the true class nformaton. Usng sub-classes can also express better the dstrbuton of smbols resultng n mproved accuraces. We nput the orgnal re-sampled and smoothed ponts (secton 2) for each canddate character pattern nto the P2DBMN-MQDF recognzer to obtan a recognton score. We use three unar class-relevant feature functons to evaluate the node attrbutes (character nner structure, character sze, sngle-character poston), and use a bnar class-relevant feature functon and a bnar class-rrelevant feature functon to capture the dependences between nodes (par-character poston, horzontal overlap). We use log-lkelhood scores usng quadratc dscrmnant functon (QDF) classfers for them Parameter Learnng We tran the weghtng parameters Λ b a genetc algorthm (GA) usng vald data to maxmze the recognton rate on vald data. We treat each one of weghts as an element of a chromosome. For evaluatng the ftness of a chromosome, each tranng word pattern s searched for the optmal path evaluated usng the weght values n the chromosome. To save computaton, we frst set each weght value as 1, and select the top 300 recognton canddates (segmentaton-recognton paths) for each tranng word. We then tran the weght parameters b GA usng the selected 300 recognton canddates. After some teratons, we use the updated weght values to re-select top 300 recognton canddates for each tranng word pattern. We repeat recognton canddate selecton three tmes. range of the number (length) of feature ponts for each character class. For nstance, the length range of feature ponts for the character O s from 5 to 50 as shown n Fg. 2. Then, the length range of feature ponts to the termnal for each node of the tre can be calculated accordng to the length range of feature ponts for each character class, where the numbers shown n parentheses of each node box are the length range of feature ponts to the termnal for the node. We can restrct the searched paths b the length ranges resultng n mproved recognton accurac. We use the length restrcton of feature ponts for the segmentaton-free method whle usng the length restrcton of nodes (characters) to the termnal for the segmentaton-based method. For nstance, the lengths of nodes to the termnal for the character O are 3, 5 and Segmentaton-free method We use the example of the extracted feature ponts of a word pattern as shown n Fg. 1, and show a search n Fg. 3 to descrbe our processes. We conduct the search and expanson from the expanson depth d 1 to d 5. We frst search the tre lexcon from ts start nodes. In Fg. 2, the start nodes are [O], [p], and [], and based on these we expand the root node and set ts chldren nodes N 1-1 wth a character categor [O], N 1-2 wth a character categor [p] and N 1-3 wth a character categor [], where each chld node has a character categor C and a start pont of a feature sequence. The share the start pont f 1. For each node, the length from the start pont f 1 to the termnal pont f 56 s 56 so that the node N 1-3 s erased because the length range to the termnal of the tre node [] s 9-44 and does not satsf the length. For each node of N 1-1 and N 1-2, we match the feature ponts from each start pont f to the pont f +M Fgure 2: Porton of tre lexcon of words. Fgure 3: Segmentaton-free search. Fgure 4: Segmentaton canddate lattce. 4. Search methods To search the optmal path as the recognton result, we attempt segmentaton-free and segmentaton-based methods. For the two methods we use a character-snchronous beam search strateg where frst a tre lexcon s constructed from a word database, as shown n Fg. 2. When tranng character MRFs, we can count the wth the states S={s 1, s 2, s 3,,s J } of the character MRF model of the correspondng C b the Vterb search, where M s the maxmum number of the feature ponts of C and t s calculated from tranng patterns. Before matchng, we shft the X coordnates of the feature ponts from f to f +M so that the mnmum of X coordnate s 0. Then we can get MRF paths at the [End] state whch correspond to some end ponts such as f 16, f 17 and f 18 for N 1-1.We sort the scores of the MRF paths (note that each score S MRF s normalzed b 227

4 the number N fea of the feature ponts from the start pont to the end pont of the path to S MRF / N fea ), and select N MRF top MRF paths. In Fg. 3, N MRF s set as three and three MRF paths (from f 1 to f 16, from f 1 to f 17, from f 1 to f 18 ) of N 1-1, and three MRF paths (from f 1 to f 12, from f 1 to f 15, from f 1 to f 18 ) of N 1-2 are selected, and we call them sub-nodes. Node N 1-1 has three sub-nodes (N , N and N ) whle N 1-2 has three sub-nodes (N , N and N ). For all the sub-nodes up to the depth d 1, we evaluate all search paths accordng to the evaluaton crteron n eq. (3), sort them, then onl select several top paths and erase others. The number of the selected top paths s called the beam band. In Fg. 3, the beam band s set as two, and two paths endng at N and N are selected for d 1. Then, we appl the same method to process the next depth. Fnall, the expanson reaches to the depth d 5. For all nodes up to the depth d 5, we evaluate all paths accordng to the path evaluaton crteron, sort them, and then select the optmal path as the recognton result. For each node N k-j wth a character categor C and a start pont f, we need to execute MRF matchng to decde ts end ponts and sub-nodes. After that we need to extract features to recognze each sub-node b the P2DBMN-MQDF recognzer. Dfferent nodes ma be requred for the same par of C and f. Therefore, for each par of C and f, once ts end ponts and sub-nodes are decded and the scores of ts sub-nodes are calculated, we store them and use them for other nodes from the second tme. For each par of start and end ponts, we also store the features for the P2DBMN-MQDF recognzer, and use them for the second pass. Ths can greatl mprove recognton tme. We call ths storage strateg of scores and features (SSSF) Segmentaton-based method The method selects some canddate segmentaton ponts P={p 1, p 2, p 3,, p q } from the feature ponts F={f 1, f 2, f 3,, f g }, and onl consders the selected ponts for segmentaton. We set the pen-up, pen-down and the mnma and Fgure 5: Segmentaton-based search. maxma ponts as canddate segmentaton ponts whch are used to over-segment the word nto prmtves. One or more consecutve prmtve segments are combned to form a canddate character pattern. All possble combnatons of these canddate character patterns are represented n a lattce where each node denotes a canddate character pattern and each edge denotes a segmentaton pont. Each canddate character pattern that exsts between segmentaton ponts p k and p l (k, l = 1~g) s followed b canddate character patterns that start from segmentaton pont p l. A canddate character pattern does not start from a pen-up pont, and does not end at a pen-down pont. We also restrct the wdth of each canddate character pattern to a threshold value. Fgure 4 shows a segmentaton canddate lattce for the word pattern as shown n Fg. 1. For each node n the lattce possble lengths to the termnal node are calculated. Ths s done b frst settng the length of all termnal nodes as one and then workng backwards one level at a tme. Thus, for each precedng node, the length vector s calculated b addng one to the length of ts succeedng nodes. Ths s shown n Fg. 4, where the numbers n each node box refer to the possble lengths. Ths length vector s used n conjuncton wth the lexcon to prune out unlkel search paths thereb mprovng both recognton accurac and speed. We show a search n Fg. 5. d 1 - d 5 represent the depth levels n the search space. Startng wth the root node, lattce nodes (1) - (6) (Fg. 4) are expanded snchronousl at depth d 1. At N 1-1 lattce node (1) can be matched to three possble characters (O, p, ) from the lexcon (Fg. 2). Snce the lengths for node (1) are 6-19 and the lexcon length from p s 4 and that from s 2 and the do not match these possble lengths, we drop them. The case s smlar for N N 1-6. Ths forms the expanson for depth d 1. At ths stage, all paths are evaluated accordng to the path evaluaton crteron n eq. (3) and a few top scorng paths are selected whle others are pruned out. The number of selected paths s called the beam wdth. In Fg. 5 the beam wdth s two and two paths endng at N 1-5 -O and N 1-6 -O are selected for d 1. Smlarl, at depth d 2 ths process s repeated to expand the search. Fnall, ths expanson s repeated for depth d 5 eldng one canddate - Offer as the recognzed word. A canddate character pattern (lattce node) ma appear n several search paths. Evaluatng node-character matches usng a character recognzer requres extracton of features from the character pattern. In order to avod redundant processng, for each node, once ts features are extracted, we store them. These are then used n subsequent calls. Addtonall, for a node and a character categor, we store the recognton score at the frst recognton and use ths stored score for subsequent evaluatons. We also call ths SSSF smlar to the segmentaton-free method. 5. Experments We evaluate the word recognzers usng a publcl avalable dataset IAM-onDB [6], whch conssts of handwrtten sentences acqured from a smart whteboard. It has 4 dsjont sets: a tranng set (5,364 lnes); two 228

5 valdaton sets (1,438 lnes and 1,518 lnes); and a test set (3,859 lnes). For our solated word recognton, we generate word-level ground truth usng a two step approach as elaborated n [7], where a few words gvng segmentaton errors were dropped, however ther numbers were far too less to make a sgnfcant mpact on the results. We set the tre lexcon to have onl one word (the word strng label of the recognzed word pattern), and used the word recognzer traned n the prevous research [4] to recognze each word pattern and set the recognzed segmentaton result as the character segmentaton labels of the pattern. After that, we obtaned character-level segmentaton labels for the word patterns of IAM-onDB. We used the tranng data to tran the feature functon models. Then, we tested the recognton rate for the testng data. Each of these models was traned on 78 smbols. The constructed tre lexcon contans 5,562 words (actual sze of the test set lexcon). The weghtng parameters Λ were estmated usng the valdaton sets b GA. The experments were mplemented on an Intel(R) Core(TM) 2 Quad CPU 2.83GHz 2.83 GHz wth 4.00 GB memor. Table 1 shows the results. The numbers enclosed n square brackets are the average recognton tme for a word. Table: Word recognton rate (%). Method Wtout ALL Wtout Wtout geometrcwthout length Perform. MRF P2DBMN-MQDF features restrctng Seg. 300-beam 85.96[1.0s] based 1000-beam 86.54[1.2s] Seg. 300-beam 85.58[2.8s] free 1000-beam 86.12[7.4s] From these results, we can see that (1) usng all feature functon models sgnfcantl mproves the recognton accurac b combnng a MRF model wth a P2DBMN-MQDF recognzer. The MRF model apples a structural method that s weak at collectng global character nformaton, but robust aganst character shape varatons. The 2DBMN-MQDF recognzer uses an un-structural method that s robust aganst noses but weak aganst character shape varatons. B combnng them, the compensate for ther respectve dsadvantages. (2) The length restrctng can brng hgher recognton accurac. (3) The two methods (segmentaton-free and segmentaton-based) elded comparable results and the segmentaton-based method performed a lttle better. Although both methods have used the same path evaluaton crteron n a CRF model, ther recognton results are dfferent. We consder the segmentaton-based method attempts to select canddate segmentaton ponts at true characters boundares, resultng n reduced confuson and mproved recognton speed. The segmentaton-free method uses the MRF model to select N MRF top MRF paths and some end ponts from a start pont for a character categor onl n one search resultng n mproved processng speed. Therefore t s possble to sgnfcantl mprove recognton speed b combnng the two methods (segmentaton-free and segmentaton-based). (4) Our sstem outperforms a state-of-the-art recurrent neural network BLSTM wth recognton rate 85.3% for the same datasets and lexcon [1]. We also evaluate the SSSF and got a result that at beam band 300 t led to greatl mproved recognton speed (b about eght tmes for segmentaton-based method, 89 tmes for segmentaton-free method). The memor consumpton of our recognzer s about 8.5MB. 6. Concluson Ths paper presented a sstem for onlne handwrtten Englsh cursve word recognton and made a comparson between the segmentaton-free and segmentaton-based methods at the same path evaluaton crteron n a CRF model. Expermental results demonstrate the segmentaton-based method performed better because t attempts to select canddate segmentaton ponts resultng n reduced confuson and mproved recognton speed. We expect to sgnfcantl mprove recognton speed b combnng the two methods n the future. We have alread shown that combnng MRF model wth P2DBMN-MQDF acheves ver hgh recognton rate for onlne handwrtten Japanese text. In ths paper, we have shown ths combnaton s also ver effectve and successful for onlne Englsh word recognton. Acknowledgement Ths research s partall beng supported b Grant-n-ad for Scentfc Research C-15K References [1] A. Graves, M. Lwck, S. Fern andez, R. Bertolam, H. Bunke, and J. Schmdhuber. A novel connectonst sstem for unconstraned handwrtng recognton. IEEE Trans PAMI, 31(5), , [2] B. Zhu and M. Nakagawa. On-lne handwrtten Japanese characters recognton usng a MRF model wth parameter optmzaton b CRF. Proc. 11th ICDAR, , [3] C.-L. Lu and X.-D. Zhou. Onlne Japanese character recognton usng trajector-based normalzaton and drecton feature extracton, Proc. 10th IWFHR, , [4] A. Shvram, B. Zhu, S. Setlur, M. Nakagawa, and V. Govndaraju. Segmentaton based onlne word recognton: A condtonal random feld drven beam search strateg. Proc. 12th ICDAR, , [5] X.-D. Zhou, D.-H. Wang, F. Tan, C.-L. Lu, and M. Nakagawa. Handwrtten Chnese/Japanese text recognton usng sem-markov condtonal random felds. IEEE Trans. PAMI, 35(10), , [6] M. Lwck and H. Bunke. IAM-OnDB-an onlne Englsh sentence database acqured from handwrtten text on a whteboard. Proc. 8th ICDAR, , 2005 [7] C. T. Nguen, B. Zhu, and M. Nakagawa. A sem-ncremental recognton method for on-lne handwrtten Englsh text. Proc. 14th ICFHR, ,

th International Conference on Document Analysis and Recognition

th 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 information

Segmentation Based Online Word Recognition: A Conditional Random Field Driven Beam Search Strategy

Segmentation 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 information

An Approach to Real-Time Recognition of Chinese Handwritten Sentences

An 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 information

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram

Shape 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 information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

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 information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning 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 information

Classifier Selection Based on Data Complexity Measures *

Classifier 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 information

Incremental MQDF Learning for Writer Adaptive Handwriting Recognition 1

Incremental 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 information

Accumulated-Recognition-Rate Normalization for Combining Multiple On/Off-Line Japanese Character Classifiers Tested on a Large Database

Accumulated-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 information

Support Vector Machines

Support 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 information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING 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 information

Signature and Lexicon Pruning Techniques

Signature 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 information

Support Vector Machines

Support 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 information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE 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 information

High-Boost Mesh Filtering for 3-D Shape Enhancement

High-Boost Mesh Filtering for 3-D Shape Enhancement Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A 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 information

Machine Learning: Algorithms and Applications

Machine 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 information

A Binarization Algorithm specialized on Document Images and Photos

A 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 information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

An Empirical Comparative Study of Online Handwriting Chinese Character Recognition:Simplified v.s.traditional

An 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 information

Smoothing Spline ANOVA for variable screening

Smoothing 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 information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge 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 information

An Optimal Algorithm for Prufer Codes *

An 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 information

UB at GeoCLEF Department of Geography Abstract

UB 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 information

Data Mining: Model Evaluation

Data Mining: Model Evaluation Data Mnng: Model Evaluaton Aprl 16, 2013 1 Issues: Evaluatng Classfcaton Methods Accurac classfer accurac: predctng class label predctor accurac: guessng value of predcted attrbutes Speed tme to construct

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying 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 information

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches Proceedngs of the Internatonal Conference on Cognton and Recognton Fuzzy Flterng Algorthms for Image Processng: Performance Evaluaton of Varous Approaches Rajoo Pandey and Umesh Ghanekar Department of

More information

Feature Reduction and Selection

Feature 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 information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. 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 information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content 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 information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua 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 information

Optimizing Document Scoring for Query Retrieval

Optimizing 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 information

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like:

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like: Self-Organzng Maps (SOM) Turgay İBRİKÇİ, PhD. Outlne Introducton Structures of SOM SOM Archtecture Neghborhoods SOM Algorthm Examples Summary 1 2 Unsupervsed Hebban Learnng US Hebban Learnng, Cntd 3 A

More information

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity

Corner-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 information

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

More information

An Image Fusion Approach Based on Segmentation Region

An 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 information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features

Recognition of Handwritten Numerals Using a Combined Classifier with Hybrid Features Recognton of Handwrtten Numerals Usng a Combned Classfer wth Hybrd Features Kyoung Mn Km 1,4, Joong Jo Park 2, Young G Song 3, In Cheol Km 1, and Chng Y. Suen 1 1 Centre for Pattern Recognton and Machne

More information

Fast Feature Value Searching for Face Detection

Fast Feature Value Searching for Face Detection Vol., No. 2 Computer and Informaton Scence Fast Feature Value Searchng for Face Detecton Yunyang Yan Department of Computer Engneerng Huayn Insttute of Technology Hua an 22300, Chna E-mal: areyyyke@63.com

More information

Straight Line Detection Based on Particle Swarm Optimization

Straight Line Detection Based on Particle Swarm Optimization Sensors & ransducers 013 b IFSA http://www.sensorsportal.com Straght Lne Detecton Based on Partcle Swarm Optmzaton Shengzhou XU, Jun IE College of computer scence, South-Central Unverst for Natonaltes,

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face 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 information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A 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 information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

Collaboratively Regularized Nearest Points for Set Based Recognition

Collaboratively 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 information

The Research of Support Vector Machine in Agricultural Data Classification

The 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 information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

A Computer Vision System for Automated Container Code Recognition

A Computer Vision System for Automated Container Code Recognition A Computer Vson System for Automated Contaner Code Recognton Hsn-Chen Chen, Chh-Ka Chen, Fu-Yu Hsu, Yu-San Ln, Yu-Te Wu, Yung-Nen Sun * Abstract Contaner code examnaton s an essental step n the contaner

More information

Pictures at an Exhibition

Pictures at an Exhibition 1 Pctures at an Exhbton Stephane Kwan and Karen Zhu Department of Electrcal Engneerng Stanford Unversty, Stanford, CA 9405 Emal: {skwan1, kyzhu}@stanford.edu Abstract An mage processng algorthm s desgned

More information

Human Face Recognition Using Generalized. Kernel Fisher Discriminant

Human 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 information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement 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 information

Foreground and Background Information in an HMM-based Method for Recognition of Isolated Characters and Numeral Strings

Foreground and Background Information in an HMM-based Method for Recognition of Isolated Characters and Numeral Strings Foreground and Background Informaton n an HMM-based Method for Recognton of Isolated Characters and Numeral Strngs Alceu de S. Brtto Jr a,b, Robert Sabourn c,d, Flavo Bortolozz a,chng Y. Suen d a Pontfíca

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM 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 information

COMPARISON 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 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 information

Load Balancing for Hex-Cell Interconnection Network

Load 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 information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Efficient Off-Line Cursive Handwriting Word Recognition

Efficient Off-Line Cursive Handwriting Word Recognition Effcent Off-Lne Cursve Handwrtng Word Recognton B. Gatos, I. Pratkaks, A.L. Kesds, S.J. Perantons To cte ths verson: B. Gatos, I. Pratkaks, A.L. Kesds, S.J. Perantons. Effcent Off-Lne Cursve Handwrtng

More information

Hierarchical clustering for gene expression data analysis

Hierarchical clustering for gene expression data analysis Herarchcal clusterng for gene expresson data analyss Gorgo Valentn e-mal: valentn@ds.unm.t Clusterng of Mcroarray Data. Clusterng of gene expresson profles (rows) => dscovery of co-regulated and functonally

More information

Correlative features for the classification of textural images

Correlative features for the classification of textural images Correlatve features for the classfcaton of textural mages M A Turkova 1 and A V Gadel 1, 1 Samara Natonal Research Unversty, Moskovskoe Shosse 34, Samara, Russa, 443086 Image Processng Systems Insttute

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL 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 information

TN348: Openlab Module - Colocalization

TN348: 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 information

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems

A 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 information

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x 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 information

A Methodology for Document Image Dewarping Techniques Performance Evaluation

A Methodology for Document Image Dewarping Techniques Performance Evaluation 2009 0 Internatonal Conference on Document Analss and Recognton A Meodolog for Document Image Dewarpng Technques Performance Evaluaton N. Stamatopoulos, B. Gatos, I. Pratkaks Computatonal Intellgence Laborator,

More information

An efficient method to build panoramic image mosaics

An efficient method to build panoramic image mosaics An effcent method to buld panoramc mage mosacs Pattern Recognton Letters vol. 4 003 Dae-Hyun Km Yong-In Yoon Jong-Soo Cho School of Electrcal Engneerng and Computer Scence Kyungpook Natonal Unv. Abstract

More information

Support Vector Machines. CS534 - Machine Learning

Support Vector Machines. CS534 - Machine Learning Support Vector Machnes CS534 - Machne Learnng Perceptron Revsted: Lnear Separators Bnar classfcaton can be veed as the task of separatng classes n feature space: b > 0 b 0 b < 0 f() sgn( b) Lnear Separators

More information

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole

The 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 information

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

Object-Based Techniques for Image Retrieval

Object-Based Techniques for Image Retrieval 54 Zhang, Gao, & Luo Chapter VII Object-Based Technques for Image Retreval Y. J. Zhang, Tsnghua Unversty, Chna Y. Y. Gao, Tsnghua Unversty, Chna Y. Luo, Tsnghua Unversty, Chna ABSTRACT To overcome the

More information

3D vector computer graphics

3D 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 information

Integrated Expression-Invariant Face Recognition with Constrained Optical Flow

Integrated Expression-Invariant Face Recognition with Constrained Optical Flow Integrated Expresson-Invarant Face Recognton wth Constraned Optcal Flow Chao-Kue Hseh, Shang-Hong La 2, and Yung-Chang Chen Department of Electrcal Engneerng, Natonal Tsng Hua Unversty, Tawan 2 Department

More information

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures A Novel Adaptve Descrptor Algorthm for Ternary Pattern Textures Fahuan Hu 1,2, Guopng Lu 1 *, Zengwen Dong 1 1.School of Mechancal & Electrcal Engneerng, Nanchang Unversty, Nanchang, 330031, Chna; 2. School

More information

Classification / Regression Support Vector Machines

Classification / 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 information

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System Fuzzy Modelng of the Complexty vs. Accuracy Trade-off n a Sequental Two-Stage Mult-Classfer System MARK LAST 1 Department of Informaton Systems Engneerng Ben-Guron Unversty of the Negev Beer-Sheva 84105

More information

Robust Vanishing Point Detection for MobileCam-Based Documents

Robust Vanishing Point Detection for MobileCam-Based Documents 011 Internatonal Conference on Document Analss and Recognton Robust Vanshng Pont Detecton for MobleCam-Based Documents Xu-Cheng Yn, Hong-We Hao Department of Computer Scence School of Computer and Communcaton

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User 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 information

Incremental Learning with Support Vector Machines and Fuzzy Set Theory

Incremental 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 information

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline mage Vsualzaton mage Vsualzaton mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and Analyss outlne mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and

More information

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION SHI-LIANG SUN, HONG-LEI SHI Department of Computer Scence and Technology, East Chna Normal Unversty 500 Dongchuan Road, Shangha 200241, P. R. Chna E-MAIL: slsun@cs.ecnu.edu.cn,

More information

CMPS 10 Introduction to Computer Science Lecture Notes

CMPS 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 information

Lecture 5: Multilayer Perceptrons

Lecture 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 information

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images

Using 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 information

Cluster Analysis of Electrical Behavior

Cluster 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 information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

Combining Multiple Resources, Evidence and Criteria for Genomic Information Retrieval

Combining Multiple Resources, Evidence and Criteria for Genomic Information Retrieval Combnng Multple Resources, Evdence and Crtera for Genomc Informaton Retreval Luo S 1, Je Lu 2 and Jame Callan 2 1 Department of Computer Scence, Purdue Unversty, West Lafayette, IN 47907, USA ls@cs.purdue.edu

More information

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation Internatonal Conference on Logstcs Engneerng, Management and Computer Scence (LEMCS 5) Maxmum Varance Combned wth Adaptve Genetc Algorthm for Infrared Image Segmentaton Huxuan Fu College of Automaton Harbn

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

Three supervised learning methods on pen digits character recognition dataset

Three supervised learning methods on pen digits character recognition dataset Three supervsed learnng methods on pen dgts character recognton dataset Chrs Flezach Department of Computer Scence and Engneerng Unversty of Calforna, San Dego San Dego, CA 92093 cflezac@cs.ucsd.edu Satoru

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining 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 information

Biostatistics 615/815

Biostatistics 615/815 The E-M Algorthm Bostatstcs 615/815 Lecture 17 Last Lecture: The Smplex Method General method for optmzaton Makes few assumptons about functon Crawls towards mnmum Some recommendatons Multple startng ponts

More information

Detection of an Object by using Principal Component Analysis

Detection 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 information

Backpropagation: In Search of Performance Parameters

Backpropagation: In Search of Performance Parameters Bacpropagaton: In Search of Performance Parameters ANIL KUMAR ENUMULAPALLY, LINGGUO BU, and KHOSROW KAIKHAH, Ph.D. Computer Scence Department Texas State Unversty-San Marcos San Marcos, TX-78666 USA ae049@txstate.edu,

More information

Histogram-Enhanced Principal Component Analysis for Face Recognition

Histogram-Enhanced Principal Component Analysis for Face Recognition Hstogram-Enhanced Prncpal Component Analyss for Face ecognton Ana-ara Sevcenco and Wu-Sheng Lu Dept. of Electrcal and Computer Engneerng Unversty of Vctora sevcenco@engr.uvc.ca, wslu@ece.uvc.ca Abstract

More information

A high precision collaborative vision measurement of gear chamfering profile

A 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 information

Machine Learning 9. week

Machine 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 information

Face Recognition Based on SVM and 2DPCA

Face 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 information

Discriminative Dictionary Learning with Pairwise Constraints

Discriminative Dictionary Learning with Pairwise Constraints Dscrmnatve Dctonary Learnng wth Parwse Constrants Humn Guo Zhuoln Jang LARRY S. DAVIS UNIVERSITY OF MARYLAND Nov. 6 th, Outlne Introducton/motvaton Dctonary Learnng Dscrmnatve Dctonary Learnng wth Parwse

More information

Feature Selection for Target Detection in SAR Images

Feature Selection for Target Detection in SAR Images Feature Selecton for Detecton n SAR Images Br Bhanu, Yngqang Ln and Shqn Wang Center for Research n Intellgent Systems Unversty of Calforna, Rversde, CA 95, USA Abstract A genetc algorthm (GA) approach

More information

A Novel Accurate Algorithm to Ellipse Fitting for Iris Boundary Using Most Iris Edges. Mohammad Reza Mohammadi 1, Abolghasem Raie 2

A Novel Accurate Algorithm to Ellipse Fitting for Iris Boundary Using Most Iris Edges. Mohammad Reza Mohammadi 1, Abolghasem Raie 2 A Novel Accurate Algorthm to Ellpse Fttng for Irs Boundar Usng Most Irs Edges Mohammad Reza Mohammad 1, Abolghasem Rae 2 1. Department of Electrcal Engneerng, Amrabr Unverst of Technolog, Iran. mrmohammad@aut.ac.r

More information

HYBRID OFF-LINE OCR FOR ISOLATED HANDWRITTEN GREEK CHARACTERS

HYBRID OFF-LINE OCR FOR ISOLATED HANDWRITTEN GREEK CHARACTERS HYBRID OFF-LINE OCR FOR ISOLATED HANDWRITTEN GREEK CHARACTERS G. amvakas,2, B. Gatos, I. ratkaks, N. Stamatopoulos,2, A. Ronots 2, S.J. erantons Computatonal Intellgence Laboratory Insttute of Informatcs

More information