Talk 4

In his talk, Professor Plamondon first introduced his research interests and the projects that are ongoing in his research group. His work is focused on developing and analyzing the modeling of trajectory perception and human movement generation, with applications to automatic processing of handwriting. This includes applications to biometry (signature verification), recognition (online handwriting processing), education (handwriting learning tools), biomedical (neuromuscular condition evaluation), and model-based document preprocessing.

The audience was then presented the definition of a stroke and its properties. Professor Plamondon insisted on the speed accuracy trade-offs, that is, the spatial accuracy and the temporal accuracy. This led to a short introduction to the kinematic theory, which states that there is a synergy between two competing systems in order to perform strokes: The agonist system, which works in the direction of the performed movement, and the antagonist system that works in the opposite direction. Knowing that the impulse response of the neuromuscular system has a lognormal profile, the combination of those competing systems leads to a delta-lognormal velocity profile, which can be modeled by seven parameters. Those parameters can be extracted from acquired delta-lognormal profiles by algorithms developed by Professor Plamondon and his team. He described the first algorithms that were developed for that purpose as well as the recently developed ones. He also presented an evaluation methodology to assess the performance of those algorithms.

Next, the effect of age on the delta-lognormal profile parameters was analyzed using rapid strokes with direction reversal acquired from old and young people. The experiment showed that the seven parameters are influenced by age. Moreover, similar changes have been observed on the agonist and antagonist components. Reporting two other experiments, he also confirmed the basic hypotheses of his model, first by highlighting the proportionality relationship between the timing of muscle activation through EMG signal analysis, and second, by pointing out the existence of a new evoked response in EEG signals that correlates to the time occurrence of the neuromuscular commands.

Professor Plamondon then presented a more general version of the kinematic theory, which takes into account the fact that the agonist and antagonist systems might not work in perfect opposition for complex movements. This generalized kinematic theory leads to a sigma-lognormal velocity profile. The parameters extraction for the sigma-lognormal model was then discussed (interactive and automatic method). Applications were also discussed in depth, like the analysis and synthesis of handwriting variability, automatic database generation, and EMG and EEG signal analysis.

He finished his presentation with an overview of the numerous other applications that may be possible, from handwriting recognition and signature verification to biomedical signal processing and the design of psychomotor tests.