Wnioskowanie jakościowe do rozpoznawania podobnych cech

• Autorzy: Jović F. , Jagnjić Z.

Warianty tytułu

EN

Qualiative reasoning for recognition of similar characters

• Języki publikacji: PL

Abstrakty

PL

Do rozpoznawania cech zaszumionych (rozmytych) zastosowano metodę wykorzystującą ich własności jakościowe. Do podobnych cech próbkowych dodawano szum o czterech stochastycznie różnych stopniach nasilenia. Szare składniki szumu binaryzowano w zakresie ziarnistości od 4*4 do 16*16. Pomiaru rozpoznawania dokonano, gdy wartość wyznaczalności przekroczyła 50%. Wyniki porównano z wynikami rozpoznawania przez trójpoziomową sieć neuronową o propagacji wstecznej szkoloną na próbce. Te same próby przeprowadzono angażując ludzi do rozpoznawania cech. Ludzie wykazali większą zdolność rozpoznawania cech niż maszyny. Sieć neuronowa i maszyna jakościowa wykazały podobną zdolność rozpoznawania cech.

EN

A method that takes qualitative features of noisy characters is used for their recognition. Sample characters were similar and the addition of noise was done in four different stocha-stically grades. The gray components of noise were binarized with different coarseness from 4*4 to 16*16. The recognition measure was taken when the value of determinacy exceeded 50%. The results were compared to the recognition feature of the sample trained 3-level back propagation neural network. The same tests were also done with the limited number of human as character recognizers. People have shown better performance than machines while neural network and qualitative machine have shown some similarities in character recognition ability.

Słowa kluczowe

PL

podobny znak; rozpoznawanie; model jakościowy; szum binarny

EN

similar character; recognition; qualitative models; binary noise

• Wydawca: Wydawnictwo Wrocławskiej Rady FSNT NOT
• Czasopismo: Inżynieria Maszyn
• Rocznik: 2000
• Tom: R. 5, z. 3/4
• Strony: 213--222
• Opis fizyczny: Bibliogr. 22 poz.

Twórcy

autor

Jović F.

• Laboratory for Artificial Intelligence, University J. J. Strossmayer in Osijek, Faculty of Electrical Engineering, Osijek

autor

Jagnjić Z.

• Laboratory for Artificial Intelligence, University J. J. Strossmayer in Osijek, Faculty of Electrical Engineering, Osijek

Bibliografia

• [1] AMIN A., Off-line arabie character recognition — the state of the art. Pattern Rec., 31(5) 1998, ss. 517-530.
• [2] CHAN K.F., YEUNG D.Y., Recognizing on-line handwriten alphanumeric characters through flexible structural machine. Pattern Rec., 32(7) 1999, ss. 1044-1114.
• [3] FOGGIA R, SANSONE C, TORTORELLA F., VENTO M., Multiclassification: reject criteria for the Bayesian combiner. Pattern Rec., 32(8) 1999, ss. 1435-1447.
• [4] HEAD H., Vigilance: a physiological state of the nervous system. Brit. J. Psychol., No. 14, 1923.
• [5] HEBB D. O., Drives and the CNS. Psychol. Rev., No. 62,1955.
• [6] JEDRA M., EL OURDIGHI A, LIMOURI M., ESSAID A., Neural nets for real time analysis of cursive writing. Displays-Technology & Appl., 19(2) 1998, ss. 77-83.
• [7] JONG J.H., HONG K.S., Binarization of noisy grayscale character images by thin line modeling. Pattern Rec., 32(5) 1999, ss. 743-752.
• [8] JOVIĆ F., Qualitative reasoning and a circular information processing algebra. Informatica, 21 1997, ss. 31-47.
• [9] KIM H.J., KIM K.H., KIM S.K., LEE J.K., On-line recognition of handwriten Chinese characters based on hidden Markov models. Pattern Rec., 30(9) 1997, ss. 1484-1500.
• [10] KIM H.J., KIM S.K., KIM K.H., LEE J.K., An HMM-based character recognition network using level building. Pattern Rec., 30(3) 1997, ss. 491-50.
• [11] LAY S.R., LEE C.H., CHENG N.J., TSENG C.C., JENG B.S., TING P.Y., WU Q.Z., DAY M.L., On-line Chinese character recognition with effective candidate radical and candidate character selection. Pattern Rec., 29(10) 1996, ss. 1647-1659.
• [12] LEE J.S., KWAN O.J., BANG S.Y., Highly accurate recognition of printed Korean characters through an improved two-stage classification process. Pattern Recognition Soc, 1999.
• [13] LIN J.R., CHEN C.F., Stroke extraction for Chinese characters using a trend-followed transcribing technique. Pattern Rec., 20(11) 1996, ss. 1785-1805.
• [14] MACKWORT N. H., Researches on the measurement of human performance, in: Selected papers on human factors in the design and use of control systems. Dover Publ., New York, 1950, 1961.
• [15] MALAVI YA A., PETERS L., Fuzzy feature description of hand writing patterns. Pattern Rec., 30(10) 1997., ss. 1591-1604.
• [16] OOMMEN B.J., LOKE R.K.S., Pattern recognition with substitutions, insertions, deletions and generalized transposition. Pattern Rec., 30(5) 1997, ss. 789-800.
• [17] PAR1C H.S., LEE S.W., A truly 2-D hidden Markov model for off-line handwriten character recognition. Pattern Rec., 31(12) 1998, ss. 1849-1864.
• [18] SHUONO H., FUKUSHIMA K., Character recognition in cursive handwriting using selective attention model with bend processing. Systems & Computers in Japan, 26(10) 1995, ss. 35-46.
• [19] SINGH S., AMIN A., Neural network recognition ofhand-writen characters. Neural Computing & Appl., 8(1) 1999, ss. 67-76.
• [20] STAROSTENKO O., NEME J.A., Automatic complex glyphs recognition and interpretation. IEICE Trans, on Fundamentals of Electronic Communications & Computer Sciences, E82A(19) 1999, ss. 2154-2160.
• [21] TSENG Y.H., KUO C.C., LEE H.J., Speeding up Chinese character recognition in an automatic document reading system. Pattern Rec., 31(15) 1998, ss. 1601-1612.
• [22] YEUNG D.S., FONG H. S., A fuzzy substroke extractor for handwriten Chinese characters. Pattem Rec., 29(12) 1999, ss. 1913-1980.