HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation

Incremental Learning Of Evolving Fuzzy Inference Systems : Application To Handwritten Gesture Recognition

Abstract : We present in this thesis a new method for the conception of evolving and customizable classification systems. The main contribution of this work is represented by proposing an incremental approach for the learning of classification models based on first-order Takagi-Sugeno (TS) fuzzy inference systems. This approach includes, on the one hand, the adaptation of linear consequences of the fuzzy rules using the recursive least-squares method, and, on the other hand, an incremental learning of the antecedent of these rules in order to modify the membership functions according to the evolution of data density in the input space. The proposed method, Evolve++, resolves the instability problems in the incremental learning of TS models thanks to a global learning paradigm in which antecedents and consequents are learned in synergy, contrary to the existing approaches where they are learned separately. The performance of our system had been demonstrated on different well-known benchmarks, with a special focus on its capacity of learning new classes. In the applicative context of handwritten gesture recognition, this system can adapt continuously with the special writing styles (personalization of existing symbols) and the new needs of the users (adding new symbols). In this specific domain, we propose a second contribution to accelerate the learning of new symbols by the automatic generation of artificial data. The generation technique is based on Sigma-lognormal theory, which proposes a new representation space of handwritten forms based on a neuromuscular modeling of writing mechanism. By applying some deformations on the Sigma-lognormal profile of a given gesture, we can obtain synthetic handwritten gestures that are realistic and close to human deformation. Integrating this data generation technique in our systems accelerates the learning process and significantly improves the overall performance.
Complete list of metadata

Cited literature [87 references]  Display  Hide  Download

Contributor : Abdullah Almousa Almaksour Connect in order to contact the contributor
Submitted on : Sunday, October 14, 2012 - 12:26:52 PM
Last modification on : Tuesday, October 19, 2021 - 11:58:55 PM
Long-term archiving on: : Saturday, December 17, 2016 - 12:58:28 AM


  • HAL Id : tel-00741574, version 1


Abdullah Almaksour. Incremental Learning Of Evolving Fuzzy Inference Systems : Application To Handwritten Gesture Recognition. Machine Learning [cs.LG]. INSA de Rennes, 2011. English. ⟨tel-00741574⟩



Record views


Files downloads