On-Line Arithmetic Based Reprogrammable Hardware Implementation of LVQ Neural Network for Alertness Classification

M. Boubaker (1) , Khaled Ben Khalifa (1) , Bernard Girau (2) , Mohamed Dogui (3) , Mohamed Hédi Bedoui (1)

1 TIM - Laboratoire Technologie et Imagerie Médicale [Monastir]
2 CORTEX - Neuromimetic intelligence
3 Service d'Exploration Fonctionnelle du Système Nerveux

The current study presents the hardware implementation of a learning Vector Quantization (LVQ) neural network. Starting from the spectral EEG analysis, we suggest an LVQ serial on-line architecture implementation on a Field programmable Gate Array (FPGA) circuit. Our concern was mainly to get a light, easy-to-wear system for the classification of vigilance levels in humans using EEG signals. The results of these classified states by LVQ mode are presented in this paper. Furthermore, the highly satisfactory performances of our implementation in terms of area speed and delay are described.

Mots clés: LVQ neural network; on-line arithmetics; FPGA; EEG and vigilance

Domaines: Réseau de neurones [cs.NE]

M. Boubaker, Khaled Ben Khalifa, Bernard Girau, Mohamed Dogui, Mohamed Hédi Bedoui. On-Line Arithmetic Based Reprogrammable Hardware Implementation of LVQ Neural Network for Alertness Classification. International Journal of Computer Science and Network Security, 2008, 8 (3), pp.260-266. ⟨inria-00338767⟩