This research has been conducted in the context of the ArtiMuse project that aims at the modeling and renewal of rare gestural knowledge and skills involved in the traditional craftsmanship and more precisely in the art of the wheel-throwing pottery. These knowledge and skills constitute the Intangible Cultural Heritage and refer to the fruit of diverse expertise founded and propagated over the centuries thanks to the ingeniousness of the gesture and the creativity of the human spirit. Nowadays, this expertise is very often threatened with disappearance because of the difficulty to resist to globalization and the fact that most of those "expertise holders" are not easily accessible due to geographical or other constraints. In this paper, a methodological framework for capturing and modeling gestural knowledge and skills in wheel-throwing pottery is proposed. It is based on capturing gestures using wireless inertial sensors and statistical modeling. In particular, we used a system that allows for online alignment of gestures using a modified Hidden Markov Model. This methodology is implemented into a Human-Computer Interface, which permits both the modeling and recognition of expert technical gestures. This system could be used to assist in the learning of these gestures by giving continuous feedback in real-time by measuring the difference between expert and learner gestures. The system has been tested and evaluated on different potters with a rare expertise, which is strongly related to their local identity.