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

Size: px
Start display at page:

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

Transcription

1 th 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 1, Masak Nakagawa 2 and Venu Govndaraju 1 1 Center for Unfed Bometrcs and Sensors, Department of Computer Scence and Engneerng, Unversty at Buffalo, NY 2 Department of Computer and Informaton Scences, Tokyo Unversty Agrculture and Technology, Tokyo, Japan {ashvram, setlur, govnd}@buffalo.edu, {zhublan,nakagawa}@cc.tuat.ac.jp Abstract In ths paper we undertake recognton of onlne unconstraned cursve handwrtten Englsh words. In contrast to popular dynamc programmng or HMM-based approaches we propose a Condtonal Random Feld (CRF) drven beam search strategy appled n a combned segmentaton-andrecognton framework. Frst, a canddate segmentaton lattce s bult usng over-segmented prmtves of the word patterns. Recognton s accomplshed by synchronously matchng lexcon words wth nodes of the lattce. Probable search paths are evaluated by ntegratng character recognton scores wth geometrc and spatal characterstcs of the handwrtten segments n a CRF (condtonal random feld) model. To make computaton effcent, we use beam search to prune the set of lkely search paths. Ths overall system has been benchmarked on a new publcly avalable dataset - IBM_UB_1 as well as on the exstng UNIPEN dataset for comparson. Keywords recognton; tre-lexcon; beam search; Condtonal Random Feld; onlne; cursve; unconstraned handwrtng (key words). I. INTRODUCTION The pervasveness of moble touch screen computng devces lke tablets and smartphones has led to a push towards more flud nteracton wth these electroncs. A natural way of enterng text s by wrtng. Ths has ushered the development of applcatons that seek to recognze electronc handwrtten content. Google s recently launched Google-handwrte feature facltates handwrtten search queres [1]; also, the German automotve manufacturer Aud equpped a few of ther models wth handwrtng-supportve on-board computers to take n user nstructons as an alternatve to pressng buttons [1]. An mportant ssue wth such applcatons s that technology should be robust to recognzng handwrtng wth all the ntrnsc stylstc ssues and nose pecular to a person s wrtng rather that requrng the user to modfy hs/her wrtng style to accommodate the system. Nether should t requre person-specfc, extensve ntal tranng. Only then can t be looked at as a natural user experence. Hence, creatng a wrter-ndependent, realtme, onlne, cursve handwrtng recognton system s necessary. Ths forms the man goal of the current research. The system we descrbe n ths paper requres mnmal preprocessng of data for stylstc normalzatons and adopts a Condtonal Random Feld drven beam search algorthm that n turn, uses a tre-structure to effcently search through lexcon entres. We buld and test ths system on a new publcly avalable dataset, namely, IBM_UB_1 [2]. Ths data was collected on the CrossPad devce whch was modeled as a notepad where users wrote on actual paper usng a specal pen. Thus, samples n ths dataset reflect the varety and complexty nvolved n real lfe handwrtng. II. MOTIVATION When dealng wth automatc recognton of wrtng, a number of ssues arse that vary n complexty and nature dependng on the mode of data capture (onlne or offlne), pattern of wrtng (dscrete, cursve, mxed), representaton of data to the recognzer (characters, sub-characters, wordlevel features etc.) and the underlyng objectve of the recognton task [3]. In the onlne mode, the man objectve s to recognze the word as the user wrtes t [4].e., real-tme recognton. Ths requres fast processng algorthms. Wth ths objectve n mnd we have taken two desgn decsons (1) we buld our lexcon usng a tre-structure that speeds up processng consderably n comparson to a flat-structure as shown n [5], and, (2) we use a beam search algorthm that prunes out lexcon paths that are not lkely to yeld good results. Ths s n contrast to past research approaches whch rely on performng an exhaustve search of all words n the lexcon. Examples of such approaches nclude the HMMbased models n [6] and [7] as also n the work of Graves et al. [8] who use a recurrent neural network (RNN) to recognze unconstraned Englsh words n sentences. Wth regard to data representaton, there are two broad approaches to recognze wrtten words. One, whch s often called the analytcal approach, treats a word as a sequence of sub-unts (such as characters, strokes etc.). It frst breaks the word down to basc unts, analyzes them, and tes them together [9]. For example, n the HMM based systems of Hu et al. [7] and Lwck et al. [6] each character s modeled as a separate HMM and lexcon word HMMs are constructed by concatenatng character HMMs. In contrast, the other approach treats a word as a whole ndvsble unt and ams to recognze t usng characterstcs of the entre word. Ths s referred to as the holstc or word-based approach [9]. The man advantage of an analytcal approach s that a large /13 $ IEEE DOI /ICDAR

2 number of words can be modeled usng a fnte set of subunts such as characters, and therefore, we opt for an oversegmentaton based algorthm that dvdes each word nto prmtve segments (character or part of a character) that are later merged n an ntegrated segmentaton-and-recognton framework. However, breakng apart a word nto canddate character segments poses sgnfcant challenges. Ths s especally the case for words wrtten ether entrely n cursve or a mxed style [4]. Thus, n order to better dentfy characters n the word pattern, we use a model-drven (Condtonal Random Feld) approach that combnes shape and geometrc features of lkely prmtves to construct possble character segments and smultaneously recognze the word. To the best of our knowledge the only other CRF mplementaton for Englsh word recognton pertans to offlne handwrtng [10]. Shetty et al. s approach also sgnfcantly dffers from ours n that they employ dynamc programmng as opposed to our tre-lexcon based beam search. Moreover, our lexcon s over an order of magntude larger n comparson (5000 vs. 300). Followng our motvaton and desgn choces, we now explan the recognton system n detal below. III. PREPROCESSING AND OVER-SEGMENTATION Gven a word, we frst normalze t to a predetermned heght and extract feature ponts usng the method descrbed n [11]. The feature ponts are so extracted that the overall shape of the word pattern as well as the pen-down and penup ponts of each stroke are retaned whle smultaneously down-samplng the number of onlne coordnate ponts substantally. Subsequently, we remove delayed strokes and detect the baselne and corpus lne of the word usng regresson lnes that approxmate the local mnma and maxma of the stroke trajectory as descrbed n [5] (Fgure 1). We then set the pen-down, pen-up and the mnma and maxma ponts as canddate segmentaton ponts P={p 1, p 2, p 3,, p g } whch are used to over-segment the word nto prmtves. that each canddate character pattern that exsts between segmentaton ponts p k and p l (k, l 1~g) s followed by canddate character patterns that start from segmentaton pont p l+1 (nstead of segmentaton pont p l ). We also do not start a canddate character pattern from a pen-down pont and do not end a canddate character pattern at a pen-up pont. Fgure 2 shows a segmentaton canddate lattce (henceforth referred to as lattce ) for a sample word. Paths n the lattce are assumed to end at a termnal node. For each node n the lattce a vector of possble lengths to the termnal node s calculated. Ths s done by frst settng the length of all canddate nodes wth the fnal prmtve segment as one and then workng backwards one level at a tme. Thus, for each precedng node, the length vector s calculated by addng one to the length of ts succeedng nodes n the lattce sequence. Ths s shown n Fgure 2, 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 unlkely search paths (.e., lexcon words wth lengths dfferent from that of the current lattce path) thereby mprovng both recognton accuracy and speed. B. Search and Recognton The second module nvolves matchng sequences n the lattce wth entres n the lexcon. For ths purpose, the lexcon s set up as a tre [5] (Fgure 3). In the lattce, the search space s expanded for each depth (or level) by synchronously matchng every node at a partcular depth to characters at a smlar depth n the tre. Ths s done by restrctng the search to only those word paths n the tre that have the same precedng path sequence as well as the same succeedng path lengths as that of the lattce node beng expanded. Each search path endng at a matched nodecharacter par (probable word strng) s evaluated accordng to a path evaluaton crteron set by the CRF model detaled n the next secton. Here, adoptng a beam search, a prespecfed beam wdth s used to prune the search paths at every depth level,.e., for a beam wdth of n, at every level the paths wth the top n scores are retaned. All other paths outsde of ths wdth are pruned away. (a) (b) Fgure 1: Preprocessng. (a) Orgnal word (b) Preprocessed word black dots denote feature ponts, red denote pen-down/up; blue and green denote maxma and mnma respectvely. IV. RECOGNITION SYSTEM A. Canddate lattce constructon The frst module of the recognton system nvolves creatng a canddate lattce for the word to be recognzed from the over-segmented prmtves. One or more consecutve prmtve segments are combned to form a canddate (probable) character pattern. All possble sequence combnatons of these canddate character pattern strngs are represented n a lattce where each node n the lattce denotes a canddate character pattern and each edge denotes a segmentaton pont leadng to the next pattern n the sequence. In order to skp lgatures, we defne an edge such Fgure 2: Segmentaton Canddate Lattce 853

3 We walk through the above descrbed charactersynchronous search procedure usng a synthetc example (Fgures 2, 3 and 4). In Fgure 4, d 1 - d 4 represent the depth levels n the search space. Startng wth the root node, we observe that lattce nodes (1), (2) and (3) (Fgure 2) are expanded synchronously at depth d 1 for possble character matches from the lexcon (Fgure 3). At N 1-1 lattce node (1) can be matched to three possble characters n, r and t. The case s smlar for N 1-2. At N 1-3 snce the length vector for node (3) accommodates a length-to-termnal of 5, 4 or 3 and the lexcon paths from t do not match these possble lengths (they possess lengths of 2 or 6), we drop these paths. Ths forms the expanson for depth d 1 where we have eght pattern-character matches n total. At ths stage, all eght paths are evaluated accordng to the path evaluaton crteron and a few top scorng paths are selected whle others are pruned out. The number of selected paths s called the beam wdth. In our synthetc example the beam wdth s two and for node (2), r s selected whereas for node (3), n s selected (hghlghted n red n Fgure 4). Smlarly, at depth d 2 ths process s repeated to expand the search. At depth d 2, followng the lattce, we note that for node (2) the possble successor nodes are (11), (12), (13), (14) and (15) (whch refer to N 2-1, N 2-2, N 2-3, N 2-4 and N 2-5 respectvely) whereas for node (3), the possble successor nodes are (16), (17), (18), (19) and (20) (whch refer to N 2-6, N 2-7, N 2-8, N 2-9 and N 2-10 respectvely). For N 2-1, the two possble character categores are a and o and both are retaned snce paths from both satsfy the length requrement of node (11). At N 2-2, the two possble lexcon characters are a and o. Here, only one satsfes the length requrement of node (12) ( o whch has the length 3). Hence, the path from a s dropped. Smlarly we process the cases for N 2-3 up to N 2-10 for depth d 2. Fnally, ths expanson s repeated for depth d 3 and d 4. At d 4, the path endng at node (35) w par s carred forward from d 3 because w s a termnal node n the lexcon (as long as ths path score falls wthn the beam wdth). Ths procedure yelds two canddates note and new for recognton. Fgure 3: Lexcon Tre It may be observed here that a partcular canddate character pattern (lattce node) may appear n several search paths (possble word sequences). Ths would requre matchng of the same node wth possble character matches from the tre for dfferent path evaluatons. Evaluatng such nodecharacter matches usng a character recognzer requres extracton of feature ponts from the pattern prmtves. In order to avod redundant processng, once a node s features are extracted, we store them. These are then used n subsequent calls. Addtonally, for a node and a character category, we store the recognton score at the frst recognton nstance and use ths stored node-character par score when called upon for subsequent evaluatons. In our experments, these steps greatly mproved recognton speed (by about eght tmes). Fgure 4: Character Synchronous Search V. CRFS FOR WORD RECOGNITION A. CRFs on Segmentatons Consder an nput word pattern X whch s oversegmented nto a sequence of prmtves usng a set of canddate segmentaton ponts P={p 1, p 2, p 3,, p g }. Relatng ths to the segmentaton lattce descrbed earler, one can see that a start-to-end path on the lattce s analogous to a probable segmentaton sequence S of X. Further, assume that Y refers to the label sequence of S. Fgure 5: Neghborhood graphs of Probable Segmentatons Thus, for a gven word, there are many probable segmentatons {S 1, S 2, S 3,,S m }, where each S ={s 1, s 2, s 3,,s n_} and s j denotes a canddate character pattern between segmentaton ponts p k and p l (k, l 1~g ). For each S we construct a neghborhood graph g (Fgure 5), such that each node denotes a canddate character pattern and each lnk 854

4 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. In our CRF framework, s j and (s j-1,s j) correspond to unary (sngle) and bnary (parwse) clques, respectvely. From the defnton of CRF, P(Y S,X) can be approxmated by an arbtrary real-valued energy functon E(X,S, ) wth clque set N and parameters as: exp( E( X, S, Λ)) P( Y S, X ) = Z( S, X ) Z( S, X ) = exp( E( X, S, Y ', Λ)) ( Y ') Snce Z(S,X), the normalzaton constant does not depend on t may be gnored f we do not requre strct probablty values. Then, the problem of fndng the best label Y, that nvolves maxmzng P(Y S,X), becomes equvalent to mnmzng the total graph energy: Y* = argmax y P(Y S,X) = argmn y E(X, S, ) (2) Utlzng only unary and bnary clques n our CRF model, the total energy functon can be defned as K k E( X, S, Λ) = λk f ( s j, s j ) ( y j 1, y j ) 1 ( s j =, s j ) S k 1 1 (3) k where f ( s, ) ( 1, ) 1 s y j y are feature functons on a bnary j j j clque (s j-1,s j), ={ k } are weghtng parameters, y j-1 and y j denote the labels of s j-1 and s j, respectvely. loss of generalty, n eq. (3) we use only bnary clques to descrbe the feature functons as we assume that they (s j-1,s j) subsume unary clques s j. Here, the total energy functon s used to measure the plausblty of and the smaller E(X,S, ) s, the larger wll P(Y S,X) be. In the word recognton task the focus now s on selectng the best path from all possble segmentatons {S 1, S 2, S 3,,S m }. Snce all paths need not be of unform length, we need to normalze the path scores. Therefore, we use the followng path evaluaton crteron to select the best path from all segmentatons-cum-recogntons. E( X, S, Λ) EC( S, X ) = N( S, X ) (4) N S, ) Where ( X denotes the number of bnary clques (length of word). B. Parameter Learnng We apply the MCE crteron [12] optmzed by stochastc gradent descent [13] to fnd the optmal parameter vector by maxmzng the dfference between the evaluaton crteron of the most confusng (S,Y) and that of the correct one: L (, X ) = (mn( EC ( S, Y, X )) EC ( S, Y, X )) MCE ( x) = (1 + e x ) 1 f t t (5) (1) where EC(S t,y t,x) and EC f (S,X) are the evaluaton crtera of the true path and of the most confusng path respectvely. C. Feature Functons In CRF, feature functons are used to capture the node attrbutes (unary) and local dependences between nodes (bnary n our case). Further, they can be broadly classfed nto class-relevant/rrelevant on the bass of dependence/ndependence on the character class. In our model we use two character recognton scores and fve geometrc feature functons. Of the two character recognton scores for each canddate pattern, one s gven by a P2DBMN-M classfer on drecton hstogram features [14] and another s gven by an MRF classfer [11]. Among the geometrc features are three unary features that capture the character structure (such as the number of down strokes), sze (heght, wdth) and ts relatve poston n the word (dstance from the center lne). The bnary feature functons measure overlap and postonal dfferences between adjacent character patterns. Though the latter s to be calculated for all possble pars of characters, we smplfy the process by clusterng the characters nto four superclasses accordng to the mean vector of ther unary poston features. Each of these geometrc features are extracted as feature vectors and transformed to log-lkelhood scores usng quadratc dscrmnant functon () classfers. The feature functons are summarzed n Table 1. TABLE I. SUMMARY OF FEATURE FUNCTIONS Type Features Classfer f 1 Unary class-rel. Character shape P2DBMN- M f 2 Unary class-rel. Character structure MRF f 3 Unary class-rel. Unary geometrc (Down stroke number and nner gap) f 4 Unary class-rel. Unary geometrc (Character sze) f 5 Unary class-rel. Unary geometrc (Sngle-character poston) f 6 Bnary class-rrel. Bnary geometrc (Par-character poston) f 7 Bnary class-rel. Bnary geometrc (horzontal overlap) VI. EXPERIMENTATION In order to buld and emprcally valdate our model we concurrently utlzed two publcly avalable datasets IBM _UB_1 and UNIPEN [15]. We chose ths concurrent approach n order to overcome the absence of true character segmentaton ponts for words n the IBM_UB_1 dataset. Further, ths also helped us benchmark aganst an already exstng and wdely adopted handwrtng dataset (UNIPEN). IBM_UB_1 conssts of a twn-folo structure where each author-specfc fle conssts of a summary-query par. The summary s a full page of unconstraned cursve wrtng whereas the query page conssts of approxmately 25 handwrtten words appearng n the summary counterpart. For ths research we used only the data from the query pages whch s at the word-level granularty. Wthn the UNIPEN dataset we found that a subset of samples n the tran_r01_v07 folder were shared across two categores solated characters (Benchmark #3) and solated words (Benchmark #6). Specfcally, the art folder wth 6 855

5 wrters, cea wth 6 wrters, ceb wth 4 wrters and, ka wth 28 wrters. By mappng coordnate nformaton from character to word categores for these samples, we were able to obtan words samples wth true character-level segmentaton labels. Usng ths data (14,691 characters, 2,127 words, 44 wrters), we traned our CRF-based framework to obtan ntal (a) weghtng parameters and (b) classfers of the geometrc feature functons. Wth ths ntalzed model and a lexcon of a sngle word (the ground truth word) we were able to deduce the approxmate segmentaton for each word n the IBM_UB_1 dataset. The creaton of character-level segmentaton ponts for words n IBM_UB_1 represents the frst phase of expermentaton. In the second phase, we retraned and rebult our model usng the updated IBM_UB_1 data. For ths we selected four pages at random from 20 wrters as testng data and used the remander for tranng. Due to the data-ntensve nature of CRF tranng, a majorty of the data was used for tranng. The tranng set conssted of 61,105 words and 355,895 characters across 62 categores (dgt, uppercase and lowercase Latn alphabet), whle the test set conssted of 1,795 words and 10,987 characters. Further, we used the above retraned model for testng on the UNIPEN subset too to provde results for comparson (Table 2). For both experments the beam band s set at Dataset TABLE II. RECOGNITION RESULTS Lexcon Sze Test Set Sze (words) (words) Recognton Rate IBM_UB_ % UNIPEN % Though we provde recognton rates for the UNIPEN dataset, t must be noted that t conssts prmarly of laboratory samples and s not reflectve of free, unconstraned handwrtng from the feld. The only other relatvely close dataset to IBM_UB_1 s the IAM-OnDB dataset. Here too, snce the dataset conssts of sentences and not ndvdual words we are unable to perform approprate benchmark tests. Extant research usng IAM-OnDB has reled upon language models for enhanced recognton performance [6, 8]. Nevertheless, our model wthout the advantage of language models compares favorably (Table 3). A second ssue to note s that n [8], whch uses RNN wth a language model, the metrc used for assessng recognton s smlar to strng matchng usng Levenshten dstance. Ths measure allows for more dfferences n characters wthn a word and thus s more forgvng than a smple bnary correct/ncorrect metrc. Our results are presented usng the more strngent correct/ncorrect metrc. TABLE III. ONLINE UNCONSTRAINED ENGLISH WORD RECOGNITION COMPARATIVE RESULTS Recognton Dataset Data Type Model Rate IBM_UB_1 Word Level CRF Beam 78.72% IAM OnDB Sentence Level HMM + LM 70.8% IAM OnDB Sentence Level RNN + LM 79% Recognton results wth dfferent combnatons of feature functons are summarzed below (Table 4). As evdent, the P2DBMN-M and MRF based features and search length restrctons contrbute sgnfcantly to overall accuracy. Method Perfor. Word rec. TABLE IV. f 1 f 2 EXPERIMENT RESULTS f 3-f 7 length restrctons Usng all feature functons rate 71.92% 72.53% 77.05% 72.70% 78.72% VII. CONCLUSION We explored a CRF-drven beam search method to recognze unconstraned cursve onlne Englsh words. Combnng a tre-lexcon wth a character-synchronous lattce search algorthm, we acheve recognton rates that compare favorably to the current state of the art. Our contrbuton s two-fold: (a) applcaton of a beam search strategy to enable effcent processng of the search space, and, (b) mergng beam search wth a CRF model that combnes both feature probablty scores and character recognton scores to mprove performance. Ths, we beleve, s a promsng avenue for future research. REFERENCES [1] K. Shabanova. Do Androds dream of handwrtng recognton?, [2] U. a. B. Center for Unfed Bometrcs and Sensors, IBM-UB Onlne and Offlne Mult-lngual Handwrtng Data Set, [3] S. Connell, and A. K. Jan, Onlne handwrtng recognton usng multple pattern class models, unpublshed, [4] C. C. Tappert, and S.-H. Cha, Englsh language handwrtng recognton nterfaces: San Francsco: Morgan Kaufmann, [5] S. Jaeger, S. Manke, J. Rechert, and A. Wabel, Onlne handwrtng recognton: the NPen++ recognzer, Internatonal Journal on Document Analyss and Recognton, vol. 3, no. 3, pp , [6] M. Lwck, and H. Bunke, HMM-based on-lne recognton of handwrtten whteboard notes, Fronters n Handwrtng Recognton, Proc 10 th Int. Workshop on, pp , [7] J. Hu, M. K. Brown, and W. Turn, HMM based onlne handwrtng recognton, Pattern Analyss and Machne Intellgence, IEEE Transactons on, vol. 18, no. 10, pp , [8] A. Graves, S. Fernández, M. Lwck, H. Bunke, and J. Schmdhuber, Unconstraned onlne handwrtng recognton wth recurrent neural networks, Adv. n Neural Informaton Processng Systems, vol. 20, pp. 1-8, [9] S. Madhvanath, and V. Govndaraju, The role of holstc paradgms n handwrtten word recognton, Pattern Analyss and Machne Intellgence, IEEE Transactons on, vol. 23, no. 2, pp , [10] S. Shetty, H. Srnvasan, and S. Srhar, Handwrtten word recognton usng condtonal random felds. Proc. 9 th Int. Conf. Document Analyss and Recognton, pp , [11] B. Zhu, and M. Nakagawa, A MRF Model wth parameter optmzaton by CRF for on-lne recognton of handwrtten Japanese characters, Proc. Doc. Recognton and Retreval XVIII, pp. 1-10, [12] B.-H. Juang and S. Katagr, Dscrmnatve learnng for mnmum error classfcaton, IEEE Trans. Sgnal Processng, 40(12), pp , [13] H. Robbns and S. Monro, A stochastc approxmaton method, Ann. Math. Stat. 22, pp , [14] C.-L. Lu and X.-D. Zhou, Onlne Japanese Character Recognton Usng Trajectory-based Normalzaton and Drecton Feature Extracton, Fronters n Handwrtng Recognton, 10th Int. Workshop on, pp , [15] I. Guyon, L. Schomaker, R. Plamondon, M. Lberman, and S. Janet, UNIPEN project of on-lne data exchange and recognzer benchmarks. Pattern Recognton, Computer Vson & Image Processng., Proc. of the 12th IAPR Int. Conf. on, vol.2., pp ,

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

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

More information

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

PRÉSENTATIONS DE PROJETS

PRÉSENTATIONS 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 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

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

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

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

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

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem 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 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

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

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

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

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

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

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

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. 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 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

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 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

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

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

Learning-Based Top-N Selection Query Evaluation over Relational Databases

Learning-Based Top-N Selection Query Evaluation over Relational Databases Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **

More information

Assembler. Building a Modern Computer From First Principles.

Assembler. Building a Modern Computer From First Principles. Assembler Buldng a Modern Computer From Frst Prncples www.nand2tetrs.org Elements of Computng Systems, Nsan & Schocken, MIT Press, www.nand2tetrs.org, Chapter 6: Assembler slde Where we are at: Human Thought

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

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

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew 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 information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

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

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS P.G. Demdov Yaroslavl State Unversty Anatoly Ntn, Vladmr Khryashchev, Olga Stepanova, Igor Kostern EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS Yaroslavl, 2015 Eye

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

A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION

A 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 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

The Codesign Challenge

The 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 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

Network Intrusion Detection Based on PSO-SVM

Network Intrusion Detection Based on PSO-SVM TELKOMNIKA Indonesan Journal of Electrcal Engneerng Vol.1, No., February 014, pp. 150 ~ 1508 DOI: http://dx.do.org/10.11591/telkomnka.v1.386 150 Network Intruson Detecton Based on PSO-SVM Changsheng Xang*

More information

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

Type-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 information

Writer Identification using a Deep Neural Network

Writer Identification using a Deep Neural Network Wrter Identfcaton usng a Deep Neural Network Jun Chu and Sargur Srhar Department of Computer Scence and Engneerng Unversty at Buffalo, The State Unversty of New York Buffalo, NY 1469, USA {jchu6, srhar}@buffalo.edu

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

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

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

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 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 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

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

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

D. Barbuzzi* and G. Pirlo

D. 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 information

Active Contours/Snakes

Active Contours/Snakes Actve Contours/Snakes Erkut Erdem Acknowledgement: The sldes are adapted from the sldes prepared by K. Grauman of Unversty of Texas at Austn Fttng: Edges vs. boundares Edges useful sgnal to ndcate occludng

More information

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn and

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

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

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

Decision Strategies for Rating Objects in Knowledge-Shared Research Networks

Decision Strategies for Rating Objects in Knowledge-Shared Research Networks Decson Strateges for Ratng Objects n Knowledge-Shared Research etwors ALEXADRA GRACHAROVA *, HAS-JOACHM ER **, HASSA OUR ELD ** OM SUUROE ***, HARR ARAKSE *** * nsttute of Control and System Research,

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

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps 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 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

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

Fitting: Deformable contours April 26 th, 2018

Fitting: Deformable contours April 26 th, 2018 4/6/08 Fttng: Deformable contours Aprl 6 th, 08 Yong Jae Lee UC Davs Recap so far: Groupng and Fttng Goal: move from array of pxel values (or flter outputs) to a collecton of regons, objects, and shapes.

More information

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

Face Recognition Method Based on Within-class Clustering SVM

Face Recognition Method Based on Within-class Clustering SVM Face Recognton Method Based on Wthn-class Clusterng SVM Yan Wu, Xao Yao and Yng Xa Department of Computer Scence and Engneerng Tong Unversty Shangha, Chna Abstract - A face recognton method based on Wthn-class

More information

Editorial Manager(tm) for International Journal of Pattern Recognition and

Editorial Manager(tm) for International Journal of Pattern Recognition and Artfcal Intellgence Edtoral Manager(tm) for Internatonal Journal of Pattern Recognton and Manuscrpt Draft Manuscrpt Number: Ttle: TEXT LOCALIZATION IN COMPLEX COLOR DOCUMENTS Artcle Type: Research Paper

More information

Intelligent Information Acquisition for Improved Clustering

Intelligent Information Acquisition for Improved Clustering Intellgent Informaton Acquston for Improved Clusterng Duy Vu Unversty of Texas at Austn duyvu@cs.utexas.edu Mkhal Blenko Mcrosoft Research mblenko@mcrosoft.com Prem Melvlle IBM T.J. Watson Research Center

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

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

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

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

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

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

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

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

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

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

Mercer Kernels for Object Recognition with Local Features

Mercer Kernels for Object Recognition with Local Features TR004-50, October 004, Department of Computer Scence, Dartmouth College Mercer Kernels for Object Recognton wth Local Features Swe Lyu Department of Computer Scence Dartmouth College Hanover NH 03755 A

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

Histogram of Template for Pedestrian Detection

Histogram 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 information

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15 CS434a/541a: Pattern Recognton Prof. Olga Veksler Lecture 15 Today New Topc: Unsupervsed Learnng Supervsed vs. unsupervsed learnng Unsupervsed learnng Net Tme: parametrc unsupervsed learnng Today: nonparametrc

More information

Bootstrapping Structured Page Segmentation

Bootstrapping Structured Page Segmentation Bootstrappng Structured Page Segmentaton Huanfeng Ma and Davd Doermann Laboratory for Language and Meda Processng Insttute for Advanced Computer Studes (UMIACS) Unversty of Maryland, College Park, MD {hfma,

More information

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed

More information

Some material adapted from Mohamed Younis, UMBC CMSC 611 Spr 2003 course slides Some material adapted from Hennessy & Patterson / 2003 Elsevier

Some material adapted from Mohamed Younis, UMBC CMSC 611 Spr 2003 course slides Some material adapted from Hennessy & Patterson / 2003 Elsevier Some materal adapted from Mohamed Youns, UMBC CMSC 611 Spr 2003 course sldes Some materal adapted from Hennessy & Patterson / 2003 Elsever Scence Performance = 1 Executon tme Speedup = Performance (B)

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

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

Load-Balanced Anycast Routing

Load-Balanced Anycast Routing Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance

More information

Some Advanced SPC Tools 1. Cumulative Sum Control (Cusum) Chart For the data shown in Table 9-1, the x chart can be generated.

Some Advanced SPC Tools 1. Cumulative Sum Control (Cusum) Chart For the data shown in Table 9-1, the x chart can be generated. Some Advanced SP Tools 1. umulatve Sum ontrol (usum) hart For the data shown n Table 9-1, the x chart can be generated. However, the shft taken place at sample #21 s not apparent. 92 For ths set samples,

More information

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)

Machine 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 information