Rotated　Japanese Character Recognition

We proposed a rotated character recognition method using eigen-subspace for alpha-numeric characters so far. We first construct an eigen-subspace for each category using the covariance matrix calculated from a sufficient number of rotated character patterns. Next, we can obtain a locus by projecting their rotated characters onto the eigen sub-space and interpolating between their projected points. An unknown character is also projected onto the eigen subspace of each category. A single projection and multiple projections of the input character image were proposed. Then, the verification is carried out by calculating the distance between the projected points of the unknown character and the locus. Then the multiple projections showed a higher accuracy at low dimensions than a single projection for alphanumeric 62 categories. This time, we applied it for the first class of Japanese Industrial Standard (JIS) Kanji set which includes 2,965 categories. As the result, very high recognition accuracy over 99.8% was achieved by especially multiple projections of the input rotated images