handwritten text recognition through writer adaptation - Laurent HEUTTE

handwritten cursive words. The first, called analytical, is a data-driven bottom-up approach in which letters are recognized before a lexical analysis is performed ...
128KB taille 5 téléchargements 313 vues
LQ 9 6JXUHY . %R\DQRQ DQG 0 +DGMLVNL (GV  3URFHHGLQJV RI WKH ,QWHUQDWLRQDO &RQIHUHQFH RQ $XWRPDWLFV DQG ,QIRUPDWLFV6RILD%XOJDULD0D\ -XQH

+$1':5,77(17(;75(&2*1,7,21 7+528*+:5,7(5$'$37$7,21 /+HXWWH73DTXHW$1RVDU\+.KRXIL

/DERUDWRLUH36,8QLYHUVLWpGH5RXHQ)0RQW6DLQW$LJQDQ &HGH[)UDQFH ^/DXUHQW+HXWWH7KLHUU\3DTXHW$OL1RVDU\+HOD.KRXIL`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

LQ 9 6JXUHY . %R\DQRQ DQG 0 +DGMLVNL (GV  3URFHHGLQJV RI WKH ,QWHUQDWLRQDO &RQIHUHQFH RQ $XWRPDWLFV DQG ,QIRUPDWLFV6RILD%XOJDULD0D\ -XQH EHWZHHQ WZR VXFFHVVLYH OHWWHUV /LNHZLVH WKH UHFRJQLWLRQ PRGXOH ZLOO QRW EH DEOH WR WDNH LQWR DFFRXQW D VSHOOLQJ PLVWDNH ,W ZLOO WKHUHIRUH EH QHFHVVDU\ WR DGG D OH[LFRQ YHULILFDWLRQ VWHS +RZHYHU WKH ZULWHU DGDSWDWLRQ VWUDWHJ\ ZH SURSRVH OHDQV ILUVW RQ WKH XVH RI WKH ZULWHU PRUSKRORJLFDOFRQVWUDLQWVWROLIWDPELJXLW\EHIRUH UHVRUWLQJ WR VRPHWLPHV URXJK OH[LFDO GHFLVLRQV :H SUHVHQW WKLV VWUDWHJ\LQWKHIROORZLQJVXEVHFWLRQV

WKH FODVVLILHU WKDW DOORZV WR LQWHJUDWH HDVLO\ VXFK OHDUQLQJ H[DPSOHV)LJXUHVKRZVVRPHHOHPHQWVRIWKHUHMHFWFODVV IRUHDFKOHYHORIVHJPHQWDWLRQ

6HJPHQWDWLRQDQGREVHUYDWLRQWUHLOOLV /HW 2 RR  R7 EH WKH VHTXHQFH RI HOHPHQWDU\

JUDSKHPHV REVHUYHG RQ WKH ZRUG DIWHU D ZRUG VHJPHQWDWLRQVWHS § 2 · § R R  R7 · ¸ ¨ ¸ ¨ /HW 0 ¨ 2  ¸ ¨ R R   R 7  ¸ EH WKH VHW RI ¸ ¨ 2 ¸ ¨ R R R ©  ¹ ©    7   ¹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

D

E

F )LJXUH6DPSOHVRIWKHUHMHFWFODVVE\VHJPHQWDWLRQOHYHO D VW OHYHO E QG OHYHO F UG OHYHO

7KH RYHUDOO GDWDEDVH RI JUDSKHPHV H[WUDFWHG IURP WKH VHJPHQWDWLRQ RI KDQGZULWWHQ WH[WV LQFOXGHV  HOHPHQWDU\ JUDSKHPHV OHYHO  RI VHJPHQWDWLRQ   ELJUDPV OHYHO JUDSKHPHVPHUJHG DQGWULJUDPV OHYHO    JUDSKHPHV PHUJHG  )RU HDFK OHYHO RI VHJPHQWDWLRQRIWKH RYHUDOOGDWDEDVHKDVEHHQXVHGIRU WUDLQLQJ DQG  IRU WHVWLQJ $Q DQDO\VLV RI WKH PRVW IUHTXHQW OHWWHUV IRXQG LQ WKH GDWDEDVH IRU HDFK OHYHO RI VHJPHQWDWLRQ KDV OHDG XV WR GHILQH D GLIIHUHQW QXPEHU RI OHWWHUFODVVHVSHUOHYHODVVKRZQLQ7DEOH OHYHO OHYHO OHYHO

RI FODVVHV WUDLQLQJWHVW UHMHFW          7DEOH2YHUDOOJUDSKHPHGDWDEDVH

7DEOH JLYHV WKH FODVVLILHU UHVXOWVRQ WKHWHVW VHW IRU HDFK OHYHO RI VHJPHQWDWLRQ WKH UHFRJQLWLRQ UDWHV LQFOXGH ZHOO UHFRJQL]HG OHWWHUV DQG FRUUHFWO\ UHMHFWHG JUDSKHPHV JRRG UHMHFW WKHHUURUUDWHVLQFOXGHRQRQHKDQGPLVUHFRJQL]HG OHWWHUVDQGUHMHFWHGOHWWHUVDQGRQWKHRWKHUKDQGJUDSKHPHV WRUHMHFWZKLFKKDYHEHHQUHFRJQL]HGDVOHWWHUV EDGUHMHFW   5HFRJ (UURU OHYHO   OHYHO   OHYHO   7DEOH/HWWHUUHFRJQLWLRQUDWHV

/H[LFRQPRGHO

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

LQLWLDOSUREDELOLW\YHFWRUQRWHG S L *HQHUDOO\ILUVW RU VHFRQG RUGHUVWDWLRQDU\PRGHOVDUHXVHG DQGQRWEH\RQGIRUHYLGHQWUHDVRQVRIPHPRU\FRPSOH[LW\ :H KDYH HYDOXDWHG WKH SHUWLQHQFH RI WKLV QRQ VWDWLRQDU\ PRGHO FRPSDUHG WR D FODVVLF VHFRQG RUGHU VWDWLRQDU\ PRGHO )RU WKLV FRPSDULVRQ ZH KDYH XVHG WKH SUHGLFWLYH SRZHURIWKHWUDQVLWLRQPDWULFHVRIWKHWZRPRGHOV>@)RU HDFK WUDQVLWLRQ PDWUL[ $ D L  M WKH SUHGLFWLYH SRZHU EDVHGRQ6KDQQRQHQWURS\LVGHILQHGE\

¦ ¦D 1

SSK $  

L 

ZL



M 

LM

ORJ  D LM

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

1 0.9 0.8

pph

0.7 0.6 0.5 0.4 0.3 0

5 10 position dans le mot

15

)LJXUH3UHGLFWLYHSRZHURIWKHWZRWUDQVLWLRQPRGHOV GRWWHGOLQHVWDWLRQDU\PRGHO

:ULWHUDGDSWDWLRQ 7KH ZULWHU DGDSWDWLRQ VFKHPH ZH SURSRVH LV MXVWLILHG E\ WKH H[LVWHQFH RI VRPH UHGXQGDQF\ RI HOHPHQWDU\ SDWWHUQV H[WUDFWHG IURP D KDQGZULWLQJ ZKLFK KDV OHDG XV WR WKH GHILQLWLRQ RI WKH ZULWHU LQYDULDQW FRQFHSW >@ 7KHUHIRUH WKLV FRQWH[WXDO LQIRUPDWLRQ RQ WKH VHJPHQWHG SDWWHUQV FDQ DFWDVDPRUSKRORJLFDOFRQVWUDLQWSURYLGHGWKDWRQHLVDEOH WR EUHDN WKH LQGHSHQGHQFH RI LQWHUSUHWDWLRQV HVSHFLDOO\ DW WKH JUDSKHPH UHFRJQLWLRQ OHYHO 7KLV QHHGV WKXV WR FRPSDUH WKH LQWHUSUHWDWLRQV DFFRUGLQJ WR WKH LQYDULDQW FOXVWHUVLQRUGHUWRUHLQIRUFHFRKHUHQWLQWHUSUHWDWLRQV ,Q WKH IUDPHZRUN RI RXU DQDO\WLFDO GHFLVLRQ EDVHG DSSURDFKWKLVLQIRUPDWLRQDOORZVWRXSGDWHLQWHUSUHWDWLRQV SURSRVHGE\WKHOHWWHUFODVVLILHUZKLFKSURYLGHVK\SRWKHVHV E\ DQDO\]LQJ LVRODWHG JUDSKHPHV 7KLV FRQWH[WXDO LQIRUPDWLRQ FDQ EH H[SUHVVHG DV WKH SUREDELOLW\ RI WKH LQYDULDQWFOXVWHU&L JLYHQWKHOHWWHUOM WKDWLV 3 & L  O M  ,Q WKLV FDVH WKH ZULWHU DGDSWDWLRQ FRQVLVWV LQ XSGDWLQJ WKH LQWHUSUHWDWLRQ RI HDFK JUDSKHPH E\ VXFFHVVLYH LWHUDWLRQV WKDWLVDWLWHUDWLRQN Nz WRFDOFXODWH N  N  D3 R  O    D 3 &  O L M L M N 3 RL  O M ª N   R  O    D 3 N   &  O º ¦ «D3 L M L M »¼ ¬ M ZKHUH 3  RL  O M

3 RL  O M LVWKHOLNHOLKRRGGHULYHGE\

WKHN11UHFRJQL]HU 7KH SDUDPHWHU D GHSHQGV RQ WKH VWUDWHJ\ XVHG VLQFH LW HQDEOHV WR IDYRU LQ VRPH FDVHV WKH FRQWH[WXDO LQIRUPDWLRQ RYHUWKHLQLWLDOLQIRUPDWLRQRQWKHJUDSKHPH (DFK LQYDULDQW FOXVWHU LQWHUSUHWDWLRQ LH 3 N  & L  O M  FDQEHGHULYHGDVIROORZV DWWKHFRQFOXVLRQRIWKHERWWRP XS VWHS N WKH 9LWHUEL DOJRULWKP JLYHV WKH EHVW OHWWHU VHTXHQFH K\SRWKHVLV IRU HDFK ZRUG RI WKH WH[W LW LV WKHUHIRUH SRVVLEOH WR HYDOXDWH IRU HDFK FOXVWHU &L LWV SUREDELOLW\JLYHQHDFKOHWWHUE\  FDUG & L O M 3 & L  O M FDUG O M

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x WKH GHIDXOW UXOH FRQVLVWV LQ WKH VHOHFWLRQ RI HYHU\ ZRUGV LQ WKH WH[W 7KLV FRUUHVSRQGV WR D FODVVLFDO UHOD[DWLRQ VFKHPH DQG LV SUREDEO\ WKH ZRUVW LQWHUDFWLRQUXOH x D VHFRQG UXOH FRQVLVWV LQ WKH VHOHFWLRQ RI ZRUGV RI LQFUHDVLQJ OHQJWK GXULQJ WKH UHSHDWLQJ ORRS ,QGHHG VLQFH VKRUW ZRUGV DUHNQRZQWREHOHVVQXPHURXVIRU D SDUWLFXODU GLFWLRQDU\ ZH H[SHFW WKDW OH[LFDO FRQVWUDLQWV ZLOO GLVDPELJXDWH ZRUG K\SRWKHVLV LQ DQ HIIHFWLYH ZD\ LQ WKLV FDVH 6LPLODUO\ ZH FRXOG VHOHFW ORQJZRUGVDVZHOO x D WKLUG UXOH FDQ EH LPSOHPHQWHG EDVHG RQ JUDSKLFDO FRQVWUDLQWV ,QGHHG ZH FDQ H[SHFW  WKDW ZRUGV ZLWK JUDSKHPHVWKDWEHORQJWRELJJURXSVRILQYDULDQWVZLOO EHHDVLHUWRUHFRJQL]H 7KLV JLYHV ULVH LQGHHG WR D ZLGH UDQJH RI LQWHUDFWLRQ VWUDWHJLHVWRHYDOXDWH

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

DQG PRUSKRORJLFDO OHYHOV :H KDYH SURSRVHG D JHQHUDO IUDPHZRUNWRFRPELQHWKHVHLQIRUPDWLRQV

5HIHUHQFHV  $ %HUFKWROG &KDvQHV GH 0DUNRY HW PRGqOHV GH WUDQVLWLRQ+HUPHV  3( %UDPDO & $ +LJJLQV $ FXUVLYH UHFRJQLWLRQ V\VWHP EDVHG RQ KXPDQ UHDGLQJ PRGHOV 0DFKLQH 9LVLRQ$SSOLFDWLRQ9ROSS  1 &DWDFK /HV OLVWHV RUWKRJUDSKLTXHV GH EDVH GX IUDQoDLV /2% 1DWKDQ5HFKHUFKH  -3 &UHWWH] $ VHW RI KDQGZULWLQJ IDPLOLHV VW\OH UHFRJQLWLRQ 3URF ,&'$5¶ 0RQWUHDO &DQDGD SS  & )DURX] 0 *LOORX[ DQG-0 %HUWLOOH +DQGZULWWHQ ZRUG UHFRJQLWLRQ ZLWK FRQWH[WXDO KLGGHQ PDUNRY PRGHOV3URFWK ,:)+5.RUHDSS  / +HXWWH 7 3DTXHW -9 0RUHDX