Jean-Philippe Draye

A dynamic neural network identification of electromyography and arm trajectory relationship during complex movements

The authors propose a new approach based on dynamic recurrent neural networks (DRNN) to identify, in human, the relationship between the muscle electromyographic (EMG) activity and the arm kinematics during the drawing of the figure eight using an extended arm. After learning, the DRNN simulations showed the efficiency of the model. The authors demonstrated its generalization ability to draw unlearned movements. They developed a test of its physiological plausibility by computing the error velocity vectors when small artificial lesions in the EMG signals were created. These lesion experiments demonstrated that the DRNN has identified the preferential direction of the physiological action of the studied muscles. The network also identified neural constraints such as the covariation between geometrical and kinematics parameters of the movement. This suggests that the information of raw EMG signals is largely representative of the kinematics stored in the central motor pattern. Moreover, the DRNN approach will allow one to dissociate the feedforward command (central motor pattern) and the feedback effects from muscles, skin and joints.

Dynamic recurrent neural networks: a dynamical analysis

In this paper, we explore the dynamical features of a neural network model which presents two types of adaptative parameters: the classical weights between the units and the time constants associated with each artificial neuron. The purpose of this study is to provide a strong theoretical basis for modeling and simulating dynamic recurrent neural networks. In order to achieve this, we study the effect of the statistical distribution of the weights and of the time constants on the network dynamics and we make a statistical analysis of the neural transformation. We examine the network power spectra (to draw some conclusions over the frequential behaviour of the network) and we compute the stability regions to explore the stability of the model. We show that the network is sensitive to the variations of the mean values of the weights and the time constants (because of the temporal aspects of the learned tasks). Nevertheless, our results highlight the improvements in the network dynamics due to the introduction of adaptative time constants and indicate that dynamic recurrent neural networks can bring new powerful features in the field of neural computing.

Evidence of a preprogrammed deactivation of the hamstring muscles for triggering rapid changes of posture in humans

Normal subjects were asked to make rapid flexions of the legs from a stationary initial standing posture in a self-paced mode. Because this movement implicates a rapid change in posture, questions were asked about the type of central command which must include the rupture of the erect posture and the accomplishment of the goal directed movement. Movements of the different segments of the body were recorded and analyzed using the optoelectronic ELITE system. Electromyographic (EMG) activities of 8 muscles of the lower limb on one side were recorded, rectified and integrated. The time relationships of the different EMG signals (activation or deactivation) were analyzed with respect to selected kinetic measures of the related segments of the body. In the majority of the subjects, before the movement onset, EMG events included a specific deactivation of the tonic EMG activity of the semimembranous (SM) and semitendinous (ST) muscles (time onset relative to the onset of the legs flexion: -196.9 +/- 96.4 ms and -180.5 +/- 89.7 ms, respectively). A second event was a phasic activation of the tibialis anterior (TA) muscle (time onset: -60.5 +/- 117.6 ms). Conjugate cross-correlation analysis of these EMG signals demonstrated the existence of a common coordinated strategy between the deactivation of the hamstring and the TA activation. Even though a small horizontal displacement of the head was recorded prior to leg movement, it occurred too late to induce deactivation of the hamstring muscles. These results demonstrate that for rapid legs flexion, where the gravity forces are the main source of joint angle acceleration, the deactivation of the SM and ST muscles acts in conjunction with the phasic activation of the TA. The preprogrammed deactivation of the SM and ST muscles represents the early phase of the central command to switch from the standing to the squatting posture.

Multi-joint coordination strategies for straightening up movement in humans.

Complex movement execution theoretically involves numerous biomechanical degrees of freedom, leading to the concept of redundancy. The kinematics and kinetics of rapid straightening up movement from the squatting position were analysed with the optoelectronic ELITE system in 14 subjects. We found multiple acceleration and deceleration peaks for the hip, knee and ankle joints during the early extension phase of the movement. In order to test the temporal coordination between the angular acceleration of these joints, conjugate crosscorrelation functions (CCF) between each set of two variables were calculated. We found a bimodal distribution of the maximum CCF in positive and negative values suggesting the existence of two distinct strategies, the in-phase and the out-of-phase strategy for each pair of joints. The hip and knee coordination strategies (in- or out-of-phase) were well conserved in each subject for repetitive movements. Combination of joint pair strategies was more reproducible for the hip-knee/knee-ankle pair than for the other combinations, suggesting that the straightening up strategies are organised around the knee. We conclude that mastering of the redundancy problem can be realised by using coordination strategies characterised by opposed joint acceleration patterns.

Emergence of clusters in the hidden layer of a dynamic recurrent neural network.

Abstract. The neural integrator of the oculomotor system is a privileged field for artificial neural network simulation. In this paper, we were interested in an improvement of the biologically plausible features of the Arnold-Robinson network. This improvement was done by fixing the sign of the connection weights in the network (in order to respect the biological Dales Law). We also introduced a notion of distance in the network in the form of transmission delays between its units. These modifications necessitated the introduction of a general supervisor in order to train the network to act as a leaky integrator. When examining the lateral connection weights of the hidden layer, the distribution of the weights values was found to exhibit a conspicuous structure: the high-value weights were grouped in what we call clusters. Other zones are quite flat and characterized by low-value weights. Clusters are defined as particular groups of adjoining neurons which have strong and privileged connections with another neighborhood of neurons. The clusters of the trained network are reminiscent of the small clusters or patches that have been found experimentally in the nucleus prepositus hypoglossi, where the neural integrator is located. A study was conducted to determine the conditions of emergence of these clusters in our network: they include the fixation of the weight sign, the introduction of a distance, and a convergence of the information from the hidden layer to the motoneurons. We conclude that this spontaneous emergence of clusters in artificial neural networks, performing a temporal integration, is due to computational constraints, with a restricted space of solutions. Thus, information processing could induce the emergence of iterated patterns in biological neural networks.

Improved signal processing with dynamic recurrent neural models using ARMA-like units

We have shown that dynamic recurrent neural networks with ARMA-like units can tackle the problem of complex signal processing. In some cases of very highly nonlinear processing, their use can even be inevitable. We have shown that the Pontryagin Maximum Principle (from the theory of control) helps to elegantly derive the continuous-time learning algorithms for these complex neural architectures Finally, we have presented practical biomedical application where dynamic recurrent networks exhibit, their robustness. We are currently investigating other applications in the field of mathematics (such as interpolation tasks i.e., for the forecasting of stock market value) and of engineer.