Self-supervised writer adaptation using perceptive concepts : application to on-line text recognition

Abstract

We recently designed a hand-printed text recognizer. The system is based on three set of experts respectively used to segment, classify and validate the text (with a French lexicon : 200K words). We present in this communication writer adaptation methods. The first is supervised by the user. The others are self-supervised strategies which compare classification hypothesis with lexical hypothesis and modify consequently classifier parameters. The last method increases the system accuracy and the classification speed. Experiments are presented on a large database of 90 texts (5400 words) written by 54 different writers and good recognition rates (82%) have been obtained.

Self-Supervised Learning for Robust Video Indexing

“…The experimental results show that their novel learning technique is competitive to SVMs and outperforms various approaches for hand-gesture recognition and fingertip tracking tasks. Oudot et al [6] present a selfsupervised method for writer adaptation in an online-text recognition system. In the self-supervised method, lexical results are compared with the classification hypothesis to find errors which are then used to re-estimate classifier parameters.…”

On-line writer adaptation for handwriting recognition using fuzzy inference systems

“…It is why the adaptation process proposed here is mainly inspired by the adaptation mechanism of k-nn classifiers [11,16] i.e. the LVQ principle.…”

“…Furthermore, we suppose here that the adaptation is supervised: each example is labeled correctly. This labeling is possible by asking the user to check the recognition or by using self-supervised technique as in [11].…”