An algorithm is presented for searching in handwriting databases that applies the A~* search algorithm. All the words in the handwriting database are organized in a trie-like data structure, termed the handwritten trie, where each node of the trie contains handwritten symbols. The search algorithm uses a statistical approach, namely the hidden Markov Model (HMM) as a scorign device at each step of the A~*-based search. The new algorithm overcomes difficulties incurred by previously proposed search algorithms that utilize the same underlying database organization.
展开▼
