System and Method for Building Feature strengthened based Pattern Recognition Model

摘要

The present invention performs a first-order DTW (Dynamic Time Warping) operation for identifying a pattern of a most similar representative data sequence based on representative data sequences of patterns learned for a data sequence to be analyzed, and when there are patterns similar to the identified pattern And an apparatus and method for constructing a pattern recognition model based on a feature-highlighted DTW by extracting sub-features having distinct characteristics for enhancing discrimination between the data sequences and performing second DTW operations on the DTWs A first DTW pattern recognition unit for performing a pattern recognition by applying a DTW algorithm to data sequences of patterns learned in the pattern learning unit of the feature emphasis pattern learning unit and a data sequence t when an arbitrary data sequence t is inputted; The DTW pattern recognizer recognizes patterns similar to those defined in the similar pattern set construction section A similar pattern determination unit for determining whether a turn set exists or not; if it is determined that there are similar patterns represented by a similar pattern set, reconstructing the representative data sequences into distinct sub- A second DTW pattern recognition unit for performing a second DTW pattern recognition on one data sequence and a data sequence of the t reconstructed in the same manner, and a second DTW pattern recognition unit for finally determining a pattern based on the result of pattern recognition by the second DTW pattern recognition unit And a pattern determination unit.