Personalizable Pen-Based Interface Using Lifelong Learning

Abstract

In this paper, we present a new method to design customizable self-evolving fuzzy rule-based classifiers. The presented approach combines an incremental clustering algorithm with a fuzzy adaptation method in order to learn and maintain the model. We use this method to build an evolving handwritten gesture recognition system, that can be integrated into an application to provide personalization capabilities. Experiments on an on-line gesture database were performed by considering various user personalization scenarios. The experiments show that the proposed evolving gesture recognition system continuously adapts and evolve according to new data of learned classes, and remains robust when introducing new unseen classes, at any moment during the lifelong learning process.

Exploiting on-the-fly interpretation to design technical documents in a mobile context

“…Figure 29 presents a screenshot of "SCRIPT&GO-ELECTRICAL" and gives an overview of the complexity of the documents it allows to draw as well as the variety of available symbols (33 different classes). 1 This figure also presents many other functionalities, such as annotation capacities, text field filling, etc., that have been implemented in "SCRIPT&GO-ELECTRICAL" because they are also needed by workers. As they do not involve the sketching of the document, they 1 A video is available at http://www.youtube.com/watch?v= k02wfPDQxoM.…”

Semi-customizable Gestural Commands Approach and Its Evaluation

“…In real software, the definition and the memorization of gestures associated to every command becomes a real challenge. Today the graphic gesture recognizers are stronger and more competitive; some of them can adapt and learn incrementally the user style [12] [13]. The problem comes from the side of the user, who needs to learn all the gesture commands.…”

“…We introduce at first the SCGC approach. Then we expose the incremental gesture recognizer [12] [13] we have used to conduct experiments. Finally we present a complete experimentation which involves more than twenty users in three comparative tests to evaluate the impact of SCGC approach in terms of learning and memorization of the gestures by the users.…”

“…The recognizer needs to be able to learn with very few training samples because the set of gestures will be different for each user. Input gestures are described by a reduced set of 15 features, and classification is executed by an incremental lightweight Fuzzy Inference System that requires few training samples [12] [13].…”

“…This system is designed to be integrated into applications which need customization aspect. Actually, the system can be used as an evolving recognizer, with lifelong learning and handling of new classes of gestures (see more details in [12] [13]). The reduced size of the feature set (only 15 features) is also essential to obtain better classification performance with only few training samples.…”