Script Recognition using Inhomogeneous P2DHMM and Hierarchical Search Space Reduction

P2DHMMs manage the assignment of observations to states quite well, whereas the calculation of probabilities is a great problem. To overcome some of the defects of P2DHMMs we propose the inhomogeneous P2DHMM (IP2DHMM). In contrast to some other approaches it is able to model consistently state duration and observation matrices of fixed height at no additional cost in computational load and number of parameters. At the same time it enables an efficient pruning mechanism and allows a considerable reduction in decoding time. In order to further reduce computational load, a combination scheme for multiple recognition stages using lexical tree beam search at different resolutions and with differently complex models (IP2DHMM and 1DHMM) is presented. The resulting system is evaluated on three different databases from different applicational fields of cursive script recognition, recognition of German legal amounts and recognition of city names from live mail.