Identifying the knee joint angular position under neuromuscular electrical stimulation via long short-term memory neural
- PDF / 1,681,214 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 39 Downloads / 174 Views
ORIGINAL ARTICLE
Identifying the knee joint angular position under neuromuscular electrical stimulation via long short-term memory neural networks Héber H. Arcolezi 1,2 & Willian R. B. M. Nunes 3 & Selene Cerna 1,2 & Rafael A. de Araujo 2 & Marcelo Augusto Assunção Sanches 2 & Marcelo Carvalho Minhoto Teixeira 2 & Aparecido Augusto de Carvalho 2 Received: 6 February 2020 / Accepted: 27 August 2020 # Sociedade Brasileira de Engenharia Biomedica 2020
Abstract Purpose Recurrent neural networks (RNNs) offer a promising opportunity for identifying nonlinear systems. This paper investigates the effectiveness of the long short-term memory (LSTM) RNN architecture in the specific task of identifying the knee joint angular position under neuromuscular electrical stimulation (NMES). The standard RNN model referred to as SimpleRNN and the well-known multilayer perceptron (MLP) are used for comparison purposes. Methods Data from seven healthy and two paraplegic volunteers were experimentally acquired. These data were adequately scaled, encoded using three timestep values (1, 5, and 10), and divided into training, validation, and testing sets. These models were mainly evaluated using the root mean square error (RMSE) and training time metrics. Results The three NN models demonstrated very good fitting to data for all volunteers. The LSTM presented smaller RMSE for most of the individuals. This is even more notable when using 5 and 10 timesteps achieving half and one-third of the error from MLP and half of the error from the SimpleRNN. This higher utility comes with a substantial time-utility trade-off. Conclusion The results in this paper show that the LSTM worths deeper investigation to design control-oriented models to knee joint stimulation in closed-loop systems. Even though the LSTM takes more time for training due to a more complex architecture, time and computational costs could be increased if achieving better modeling of systems. Rather than mathematically modeling this system with several unique parameters per individual, the use of NNs is encouraged in this task where there exist high nonlinearities and time-varying parameters. Keywords Neuromuscular electrical stimulation . Spinal cord injury . Long short-term memory . Knee joint . Nonlinear system identification
Introduction Context of the problem It is well-known that spinal cord injury (SCI), which may be caused by traumatic reasons (e.g., road accidents, day-to-day
* Héber H. Arcolezi [email protected]; [email protected] 1
Femto-ST Institute, University Bourgogne Franche-Comté, UBFC, CNRS, Belfort 90000, France
2
Department of Electrical Engineering, UNESP – University Estadual Paulista, Campus of Ilha Solteira, Ilha Solteira, SP 15385-000, Brazil
3
Department of Electrical Engineering, UTFPR - Federal University of Technology, Apucarana 86800, Paraná, Brazil
falls, sports injuries) or nontraumatic ones (e.g., tumors or diseases that destroy the neurological tissue of the spinal cord), can cause partial or total paralysis under the
Data Loading...
