The Decoder Performance Comparison in the Closed-Loop BMI System Based on MPC

摘要

In recent years, some emerging technologies, such as back propagation (BP) neural network and recurrent neural networks (RNN) with long-short term memory (LSTM), have been adopted in many fields. Hence, we firstly design four decoders based on the classical Wiener filter and Kalman filter, and the emerging BP neural network and RNN with LSTM, and compare the performance of these decoders offline. Secondly, considering the advantages of the model predictive control (MPC), we design an auxiliary controller based on MPC algorithm to form a closed-loop BMI system, and compare the performance of four decoders online.