Mathieu Barnachon

Research

Publications and academic work.

Research on motion capture, action recognition, gesture interpretation, and geometric reconstruction. PDF downloads, DOI links, and BibTeX files are included where available.

Journals

Peer-reviewed journal publications.

Illustration for ongoing action recognition
2014 Pattern Recognition

Ongoing Human Action Recognition with Motion Capture

Mathieu Barnachon, Saida Bouakaz, Boubakeur Boufama, Erwan Guillou

A framework for recognizing streamed actions from motion capture data, designed for early recognition of ongoing activities.

DOI PDF BibTeX
Read abstract

Ongoing human action recognition is a challenging problem with applications in surveillance, patient monitoring, and human-computer interaction. This work introduces a framework for recognizing streamed actions from motion capture data using pose histograms derived from Hausdorff distance, Bhattacharyya distance for comparison, and dynamic time warping for alignment. The method proved effective on large datasets and outperformed multiple established approaches.

2013 Pattern Recognition Letters

A Real-Time System for Motion Retrieval and Interpretation

Mathieu Barnachon, Saida Bouakaz, Boubakeur Boufama, Erwan Guillou

An exemplar-based method for real-time human motion recognition using online motion capture data and an automaton-style motion graph.

DOI PDF BibTeX
Read abstract

This paper presents a real-time recognition method based on exemplars rather than heavy intra-class learning. It formalizes recognizable motion streams as a motion graph that can classify known actions and accept new ones incrementally. The approach uses a spatio-temporal similarity metric and a skeleton-centric coordinate system to achieve fast, view-invariant recognition.

Conferences & Reports

Selected conference papers, workshops, and technical reports.

2012 ICPR

Human Actions Recognition from Streamed Motion Capture

Mathieu Barnachon, Saida Bouakaz, Boubakeur Boufama, Erwan Guillou

PDF BibTeX
Read abstract

A real-time recognition method based on histograms of action poses extracted from motion capture data, with Bhattacharyya distance and dynamic programming used to classify actions in progress.

2011 ICCBR Workshop

Intelligent Interactions Based on Motion

M. Barnachon, M. Ceccaroli, A. Cordier, E. Guillou, M. Lefevre

PDF
Read abstract

A position paper on intelligent software systems controlled by gestures, combining motion capture concerns with case-based reasoning to support adaptive feedback and assistance.

2009 ICIG

Towards Removing Ghost-Components from Visual-Hull Estimations

Brice Michoud, Erwan Guillou, Saida Bouakaz, Mathieu Barnachon, Alexandre Meyer

Read abstract

Work on detecting and removing ghost connected components from visual hull reconstructions using calibration and silhouette data, improving downstream motion capture and rendering workflows.

Capture preview for real-time movement interpretation
2012 RFIA

Interprétation de Mouvements Temps Réel

Mathieu Barnachon, Saida Bouakaz, Erwan Guillou, Boubakeur Boufama

PDF
Read abstract

A real-time motion recognition method using exemplars rather than heavy training sets, enabling fast incremental learning and view-invariant recognition from synthetic and real data.

Illustration for gesture interaction research
2012 IHMA Workshop

Interactions Intelligentes à Base de Mouvements

M. Barnachon, M. Ceccaroli, A. Cordier, E. Guillou, M. Lefevre

PDF
Read abstract

Research on gesture-driven interaction systems with a strong emphasis on usable, adaptive, real-time recognition from motion capture streams.

2012 CORESA

Reconnaissance d'Action à Partir de Capture de Mouvements

Mathieu Barnachon, Saida Bouakaz, Boubakeur Boufama, Erwan Guillou

Read abstract

A French-language publication describing efficient, ongoing action recognition from motion capture, designed for real-time use without large training datasets.

2009 ORASIS

Reconstruction géométrique par estimation de posture

Mathieu Barnachon, Brice Michoud, Erwan Guillou, Saida Bouakaz

Read abstract

A real-time incremental approach that uses motion estimation to remove ghost geometry from shape-from-silhouette reconstructions and improve precision.

2008 AFIG

Reconstruction Géométrique et Photométrique Incrémentale d'un Personnage en Mouvement à partir de Séquences Vidéo

Mathieu Barnachon, Erwan Guillou, Brice Michoud

Read abstract

Work on incremental geometric and photometric reconstruction of a moving person from video sequences, combining voxel-based reconstruction, surface extraction, and texturing.

Thesis

Doctoral dissertation.

2013 Ph.D. Thesis

Real time action recognition from examplars

Mathieu Barnachon

A dissertation on real-time action recognition methods built around exemplar paradigms, automata, and invariant sequence analysis.

PDF BibTeX
Read summary

The thesis studies action recognition under real-time, precision, and extensibility constraints. It first examines articulation trajectories to decompose actions into atomic motion units that can be evaluated as automaton states. It then addresses robustness through invariant pose sequences and histogram-based action representations, extending integral histograms with a dynamic-programming alignment process inspired by Dynamic Time Warping.