A multi-classifier combination strategy for the recognition of handwritten cursive words

Abstract

MATRA CAP S ystemes-Departement Lecture Automatique 3,av.duCentre-78052 Saint-Quentin-en-Yvelines -AbstractThe paper describes a recognition scheme for reading handwritten cursive words using three word recognition techniques. It particularly focuses on the implementation used to combine the three techniques based on a comparative sru& of different strategies. The first holistic recognition technique derives a global encoding of the word. The other techniques both rely on the segmentatiorr of the word into letters, bur diger in the character ClassiJier they use. The former runs a statistical linear classifier, and the latter runs a neural network with a different representation of the input data. The testing, comparison, and combination studies have been perfornied on word images from mail provided by the USPS. The top choice recognition rates achieved so far correspond to 88 %, 76 %, 6.5 % with respect to lexicon sizes of 10, 100, and IO00 words.

Applying Representative Uninorms for Phonetic Classifier Combination

“…This may account for the interest in classifier combination in several areas of Artificial Intelligence (e.g. [17,22,2]), and also in speech recognition [10].…”

Word shape analysis for a hybrid recognition system

“…Ascenders/descenders and vertical bars are related to local vertical extrema used by other researchers (e.g. [4,6,9]). The use of vertical bars is also reported in an off-line recognition system [21].…”

“…Their decisions are frequently complementary. Combining results of multiple recognizers provides almost universal improvement [2,3,4,5]. This is even the case when identical vectors of features are used [3].…”

Multi-level arabic handwritten words recognition

“…In the literature, we find a thorough research works treating combination of classifiers in case of character recognition [4][5]. However research works treating the strategy of combining classifiers for word recognition are less frequent [6][7]. In our application, we find in the combination of different classifiers a solution to extract complementary information.…”