Pattern matching system

Claim 2

A pattern matching system for comparing two patterns A and B, said system comprising: first memory means (1) for storing a series of representative vectors {a(1), - - - , a(k), - - - a(k)} and a series of extraction points {P(1), - - - , P(k), - - - , P(k)}, said series {a(1), - - - , a(k), - - - a(k)} indicating connecting points of a compressed pattern which is obtained by approximating said pattern A in a piecewise straight line approximation, said series {P(1), - - - , P(k), - - - , P(k)} indicating said connecting points; second memory means (2) for storing the second pattern B {b(1), . . . , b(l), . . . b(L)} expressed as a series of feature vectors; distance calculating means (3) for calculating a distance d(l,k) between a segment connecting representative vectors a(k) and a(k-1) and the feature vector b(l) in accordance with a control address (l,k); weighting coefficient calculating means (4) for calculating a weighting coefficient depending on a distance ΔP(k)=P(k)-P(k-1); weighted distance calculating means (5) for calculating a weighted distance defined by said distance d(l, k) and said weighting coefficient; and matching means (6) for calculating a composite distance between said patterns A and B by considering said weighted distance through shifting said control address signal (l, k) from k=1, . . . , K to l=1 . . . , L, said composite distance being a measure of similarity between patterns A and B.

Claim 3

A pattern matching system comparing two patterns A and B, said system comprising: first memory means (101) for storing a series of representative vectors {a(1), - - - , a(i), - - - a(I)} and a series of extraction points {P(1), - - - , P(i), - - - , P(I)}, the series {a(1), - - - , a(i), - - - , a(I)} indicating connecting points of a compressed pattern which is obtained by approximating said pattern A in a piecewise straight line approximation, the series {P(1), - - - , P(i), - - - , P(I)} indicating said connecting points; second memory means (102) for storing a series of representative vectors {b(1), - - - , b(j), - - - , b(J)} and a series of extraction points {q(1), - - - , q(j), - - - , q(J)}, the series {b(1), - - - , b(j), - - - , b(J)} indicating connecting points of a compressed pattern which is obtained by approximating said pattern B in a piecewise straight line approximation, the series {q(1), - - - , q(j), - - - , q(J)} indicating said connection points; first distance calculating means (303) for calculating a distance d A (i, j) between a segment connecting representative vectors a (i+1) and a(i) and the point corresponding to the representative vector b(j); second distance calculating means (304) for calculating a distance d B (i, j) between a segment connecting representative vectors b(j+1) and b(j) and the point corresponding to the representative vector a(i); distance storing means (104) for storing a predetermined number of distance information, said information being composed by said distance d A (i, j) assigned to a side connecting points (i, j) and i+1, j) on a lattice of time point I×J and by said distance d B (i, j) assigned to a side connecting points (i, j) and (i, j+1) of the said lattice, said storing means being controlled responsive to the control addresses i and j; integral value storing means (106) for storing a predetermined number of integral values g, said integral values corresponding to points and sides of said lattice; recurrence formula calculating means (901, 902, 903, 904, 905, 906 and 911) for calculating a weighted sum of the integral value g previously stored and the distances d A (i, j) and d B (i, j) in a plurality of combinations between the integral value g and distances d A and d B ; and matching means (907, 908, 909 and 910) for determining a composite distance between said patterns A and B by obtaining the minimum value of said integral value g by shifting the control address i and j from i=1 and j=1 to i=I and j=J, said composite distance being a measure of similarity between patterns A and B.

Claim 4

A pattern matching system comparing two patterns A and B, said system comprising: first memory means (101) for storing a series of representative vectors {a(1), - - - , a(i), - - - a(I)} and a series of extraction points {P(1), - - - , P(i), - - - , P(I)}, the series {a(1), - - - , a(i), - - - , a(I)} indicating connecting points of a compressed pattern which is obtained by approximating said pattern A in a piecewise straight line approximation, the series {P(1), - - - , P(i), - - - , P(I)} indicating said connecting points; second memory means (102) for storing a series of representative vectors {b(1), - - - , b(j), - - - , b(J)} and a series of extraction points {q(1), - - - , q(j), - - - , q(J)}, the series {b(1), - - - , b(j), - - - , b(J)} indicating connecting points of a compressed pattern which is obtained by approximating said pattern B in a piecewise straight line approximation, the series {q(1), - - - , q(j), - - - , q(J)} indicating said connection points; first distance calculating means (303) for calculating a distance d A (i, j) between a segment connecting representative vectors a (i+1) and a(i) and the point corresponding to the representative vector b(j); second distance calculating means (304) for calculating a distance d B (i, j) between a segment connecting representative vectors b(j+1) and b(j) and the point corresponding to the representative vector a(i); distance storing means (104) for storing a predetermined number of distance information, said information being composed by said distance d A (i, j) assigned to a side connecting points (i, j) and i+1, j) on a lattice of time point I×J and by said distance d B (i, j) assigned to a side connecting points (i, j) and (i, j+1) of the said lattice, said storing means being controlled responsive to the control addresses i and j; integral value storing means (106) for storing a predetermined number of integral values g, said integral values corresponding to points and sides of said lattice; recurrence formula calculating means (901, 902, 903, 904, 905, 906 and 911) for calculating a weighted sum of the integral value g previously stored and the distances d A (i, j) and d B (i, j) in a plurality of combinations between the integral value g and distances d A and d B ; and matching means (907, 908, 909 and 910) for determining a composite distance between said patterns A and B by obtaining the minimum value of said integral value g by shifting the control address i and j from i=1 and j=1 to i=I and j=J, said composite distance being a measure of similarity between patterns A and B.