The errors are also related to the shortcomings associated with the Markovian modeling of the handwriting. The first shortcoming is the susceptibility to modeling inaccuracies
that are associated with HMM character models. It is often the case that local mismatches between the handwriting and HMM model can have a negative effect on the accumulated score used for making a global decision. The second shortcoming is the limitation of the HMMs to model the handwriting signal: The assumption that neighboring observations are conditionally independent prevents an HMM from taking full advantage of the correlation that exists among the observations of a character [30].