METHOD AND APPARATUS FOR RECOGNITION OF HANDWRITTEN CHARACTER

    公开(公告)号:JPH09128490A

    公开(公告)日:1997-05-16

    申请号:JP28591296

    申请日:1996-10-09

    Abstract: PROBLEM TO BE SOLVED: To provide a segment-oriented method for improving the trade-off between efficiency and precision for the handwritten character recognition system. SOLUTION: Shape information which is usable only through segment-like modeling is held by modeling selected segments, and the nonefficiency that all possible segments are modeled is evaded. A gap feature is defined which discriminates the spatial distance between successive strokes larger than a certain threshold value from the spatial distance between successive strokes smaller than the threshold value. A penalty evaluation point is added to an opposition hypothesis evaluation point showing a successive stroke where the spatial distance is larger than the threshold value to shift opposition hypothesis cumulative evaluation points toward the evaluation point showing that two successive strokes are discontinuous. A penalty is imposed on the opposition hypothesis evaluation point showing two successive strokes are continuous where the spatial distance is less than the threshold value.

    METHOD FOR DECIDING CHARACTER INFORMATION FROM SIGNAL AND SYSTEM FOR READING BAR CODE LABEL

    公开(公告)号:JPH1091714A

    公开(公告)日:1998-04-10

    申请号:JP19893397

    申请日:1997-07-24

    Abstract: PROBLEM TO BE SOLVED: To reduce the inter-code interference independent of the S/N and also to reduce other types of blurs by adapting a scanned signal model to the measured or detected data signal by means of the deterministic expected value maximization algorithm and then decoding the data. SOLUTION: A label is scanned by a proper scanner (4), and the back scattered light y (t) sent from the label is measured by a photodetector which produces the electric signals based on the intensity of the measured light (5). Then the characteristic information included in the measurement data signal y (t) is decoded and decided (6). At the same time, a scanned signal model j (t, θ) is adapted to the measurement data signal y (t) by means of the deterministic expected value maximization algorithm (7). Then the encoded label data are translated from the scanned signal model j (t, θ) and the measurement data signal y (t) is decoded (8).

Patent Agency Ranking