From Neuronal cost-based metrics towards sparse coded signals classification

Anthony Mouraud (1) , Quentin Barthélemy (2, 1) , Aurélien Mayoue (1) , Anthony Larue (1) , Cedric Gouy-Pailler (1) , Hélène Paugam-Moisy (3)

1 LOAD (CEA, LIST) - Laboratoire Outils d'Analyse des Données (CEA, LIST)
2 GIPSA-SIGMAPHY - GIPSA - Signal Images Physique
3 LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information

Sparse signal decomposition are keys to efficient compression, storage and denoising, but they lack appropriate methods to exploit this sparsity for a classification purpose. Sparse coding methods based on dictionary learning may result in spikegrams, a sparse and temporal representation of signals by a raster of kernel occurrence through time. This paper proposes a method for coupling spike train cost based metrics (from neuroscience) with a spikegram sparse decompositions for clustering multivariate signals. Experiments on character trajectories, recorded by sensors from natural handwriting, prove the validity of the approach, compared with currently available classification performance in literature.

Mots clés: Sparse signal decomposition; sparse coding; spikegram sparse decomposition; spike train cost based metrics; spectrum analysis; online learning; artificial intelligence; machine learning; dictionary learning; classification; clustering; signal processing; character trajectories; sparsity; clustering multivariate signals

Domaines: Traitement du signal et de l'image [eess.SP]; Mathématique discrète [cs.DM]

Anthony Mouraud, Quentin Barthélemy, Aurélien Mayoue, Anthony Larue, Cedric Gouy-Pailler, et al.. From Neuronal cost-based metrics towards sparse coded signals classification. ESANN 2012 - 20th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Apr 2012, Bruges, Belgium. pp.311-316. ⟨hal-00731497⟩