This work aims at introducing a novel way of controlling tremor in ballistic planar movements of the upper limb using Artificial Neural Networks. The idea is that of designing a smart controller based on Artificial Neural Networks, able to drive adaptively an electrical stimulator device in order to promote the voluntary movement and to deploy the tremor simultaneously. This paper would like to present the preliminary results of the feasibility study carried on by studying the neural controller when dealing with simulated movements affected by Parkinson's action tremor. A biomechanical arm model with two pairs of muscles modelled using Hill's lump circuit has been developed in order to simulate ballistic planar movement of the upper limb. The tremor has been simulated by adding noise in frequency range of Parkinson's action tremor to the torques at the joints. The neural activations that generate this noise have been found using an Artificial Neural Network trained by using a Reinforcement Learning paradigm. A Noise-Box has been developed in order to combine the neural activations of the noise at the joints with those of a correct movement, thus obtaining the neural activations of a voluntary movement affected by tremor. The Noise-Box also calculates the additive neural activations the controller has to provide to the biomechanical arm model in order to correct the tremor during the movement. Finally the controller has been implemented using a Feed Forward Artificial Neural Network receiving as inputs only the neural activations of a tremorous movement and providing as output the correcting activations. This Neural Network has been trained with a supervised paradigm using the information given by the Noise-Box. Various tests have been performed with different noise amplitudes.

This work aims at introducing a novel way of controlling tremor in ballistic planar movements of the upper limb using Artificial Neural Networks. The idea is that of designing a smart controller based on Artificial Neural Networks, able to drive adaptively an electrical stimulator device in order to promote the voluntary movement and to deploy the tremor simultaneously. This paper would like to present the preliminary results of the feasibility study carried on by studying the neural controller when dealing with simulated movements affected by Parkinson's action tremor. A biomechanical arm model with two pairs of muscles modelled using Hill's lump circuit has been developed in order to simulate ballistic planar movement of the upper limb. The tremor has been simulated by adding noise in frequency range of Parkinson's action tremor to the torques at the joints. The neural activations that generate this noise have been found using an Artificial Neural Network trained by using a Reinforcement Learning paradigm. A Noise-Box has been developed in order to combine the neural activations of the noise at the joints with those of a correct movement, thus obtaining the neural activations of a voluntary movement affected by tremor. The Noise-Box also calculates the additive neural activations the controller has to provide to the biomechanical arm model in order to correct the tremor during the movement. Finally the controller has been implemented using a Feed Forward Artificial Neural Network receiving as inputs only the neural activations of a tremorous movement and providing as output the correcting activations. This Neural Network has been trained with a supervised paradigm using the information given by the Noise-Box. Various tests have been performed with different noise amplitudes.

Severini, G., Conforto, S., Bernabucci, I., Schmid, M., & D'Alessio, T. (2008). Tremor control during movement of the upper limb using artificial neural networks. In IFMBE Proceedings - 4th European Congress of the International Federation for Medical and Biological Engineering (pp.72-75) [10.1007/978-3-540-89208-3_19].

Tremor control during movement of the upper limb using artificial neural networks

SEVERINI, GIACOMO;CONFORTO, SILVIA;BERNABUCCI, IVAN;SCHMID, Maurizio;D'ALESSIO, Tommaso
2008

Abstract

This work aims at introducing a novel way of controlling tremor in ballistic planar movements of the upper limb using Artificial Neural Networks. The idea is that of designing a smart controller based on Artificial Neural Networks, able to drive adaptively an electrical stimulator device in order to promote the voluntary movement and to deploy the tremor simultaneously. This paper would like to present the preliminary results of the feasibility study carried on by studying the neural controller when dealing with simulated movements affected by Parkinson's action tremor. A biomechanical arm model with two pairs of muscles modelled using Hill's lump circuit has been developed in order to simulate ballistic planar movement of the upper limb. The tremor has been simulated by adding noise in frequency range of Parkinson's action tremor to the torques at the joints. The neural activations that generate this noise have been found using an Artificial Neural Network trained by using a Reinforcement Learning paradigm. A Noise-Box has been developed in order to combine the neural activations of the noise at the joints with those of a correct movement, thus obtaining the neural activations of a voluntary movement affected by tremor. The Noise-Box also calculates the additive neural activations the controller has to provide to the biomechanical arm model in order to correct the tremor during the movement. Finally the controller has been implemented using a Feed Forward Artificial Neural Network receiving as inputs only the neural activations of a tremorous movement and providing as output the correcting activations. This Neural Network has been trained with a supervised paradigm using the information given by the Noise-Box. Various tests have been performed with different noise amplitudes.
This work aims at introducing a novel way of controlling tremor in ballistic planar movements of the upper limb using Artificial Neural Networks. The idea is that of designing a smart controller based on Artificial Neural Networks, able to drive adaptively an electrical stimulator device in order to promote the voluntary movement and to deploy the tremor simultaneously. This paper would like to present the preliminary results of the feasibility study carried on by studying the neural controller when dealing with simulated movements affected by Parkinson's action tremor. A biomechanical arm model with two pairs of muscles modelled using Hill's lump circuit has been developed in order to simulate ballistic planar movement of the upper limb. The tremor has been simulated by adding noise in frequency range of Parkinson's action tremor to the torques at the joints. The neural activations that generate this noise have been found using an Artificial Neural Network trained by using a Reinforcement Learning paradigm. A Noise-Box has been developed in order to combine the neural activations of the noise at the joints with those of a correct movement, thus obtaining the neural activations of a voluntary movement affected by tremor. The Noise-Box also calculates the additive neural activations the controller has to provide to the biomechanical arm model in order to correct the tremor during the movement. Finally the controller has been implemented using a Feed Forward Artificial Neural Network receiving as inputs only the neural activations of a tremorous movement and providing as output the correcting activations. This Neural Network has been trained with a supervised paradigm using the information given by the Noise-Box. Various tests have been performed with different noise amplitudes.
Severini, G., Conforto, S., Bernabucci, I., Schmid, M., & D'Alessio, T. (2008). Tremor control during movement of the upper limb using artificial neural networks. In IFMBE Proceedings - 4th European Congress of the International Federation for Medical and Biological Engineering (pp.72-75) [10.1007/978-3-540-89208-3_19].
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11590/158265
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? 1
social impact