Learning Residual Kinematic Corrections for Continuous Neural Decoding via Reinforcement Learning

Source

Source name: arXiv
Source kind: paper
Authors / organization: Jiamian Li, Niall McShane, Attila Korik, Naomi du Bois, Karl McCreadie, Leen Jabban, Benjamin Metcalfe, Özgür Şimşek, Damien Coyle
Venue: arXiv: cs.AI
Published: 2026-07-13
Collected: 2026-07-14
URL: http://arxiv.org/abs/2607.11530v1
PDF: https://arxiv.org/pdf/2607.11530v1
DOI: UNKNOWN

Plain-Language Brief

Researchers proposed a new method for improving the accuracy of decoding continuous three-dimensional motor imagery signals using non-invasive EEG-based BCIs. They introduced a two-stage decoding framework that applies reinforcement learning to correct systematic residual errors in predicted trajectories, which resulted in improved correlation and reduced Root Mean Square Errors compared to traditional methods.

What the Document Says

The research focuses on improving the accuracy of continuous 3D motor imagery decoding using EEG-based BCIs due to signal variability and residual errors.
CNN--LSTM models are used for capturing spatial and temporal dynamics, but systematic residual errors persist in predicted trajectories.
A two-stage decoding framework is proposed that uses reinforcement learning (RL) for residual kinematic correction on the outputs of a CNN--LSTM decoder.
The RL agent is trained offline without direct EEG input, instead operating on predicted kinematic trajectories to optimize movement accuracy.
Decoding performance is measured using Pearson correlation coefficients and Root Mean Square Errors along three axes: x, y, and z.
CNN--LSTM--RL improved mean correlations compared to CNN--LSTM in 2D (from 0.5076 to 0.7181) and VR (from 0.6420 to 0.7780), with relative gains of 41.5% and 21.2%, respectively.
Correspondingly, Root Mean Square Errors were reduced in 2D (from 0.0890 to 0.0532) and VR (from 0.0714 to 0.0441), representing relative reductions of 40.2% and 38.2%.
The proposed method enhances 3D BCI MI decoding by correcting kinematic errors via offline residual RL without extra neural data.

Key Concepts Mentioned

Continuous three-dimensional motor imagery
EEG-based BCIs
Convolutional neural network--long short-term memory (CNN--LSTM)
Reinforcement learning (RL)
Residual kinematic correction
Pearson correlation coefficients
Root Mean Square Errors
Virtual Reality (VR)

Claims Made by the Document

The proposed two-stage decoding framework enhances 3D BCI MI decoding by correcting kinematic errors via offline residual RL without extra neural data.

Evidence Used

Improved mean correlations and reduced Root Mean Square Errors compared to CNN--LSTM alone.
Relative gains of 41.5% and 21.2% in correlation, and relative reductions of 40.2% and 38.2% in Root Mean Square Errors.

Methods / Approach

Use of CNN--LSTM for kinematic decoding
Introduction of a two-stage decoding framework with reinforcement learning (RL) for residual kinematic correction.
Offline training of the RL agent without direct EEG input.

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

The research does not discuss potential limitations of the proposed method, such as generalizability to different datasets or subjects.

Open Questions

How well does this method perform on other datasets or with different subjects?
What is the impact of training duration and RL hyperparameters on decoding performance?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The research demonstrates an improvement in correlation and reduction in Root Mean Square Errors using the proposed two-stage decoding framework, but more studies are needed to evaluate its generalizability and optimize its performance.
Further research is required to explore the impact of training duration, RL hyperparameters, and potential limitations on decoding performance.

Raw Abstract / Excerpt

Decoding continuous three-dimensional (3D) motor imagery (MI) using non-invasive electroencephalography (EEG)-based brain--computer interfaces (BCIs) remains challenging due to signal variability and residual decoding errors. Deep learning architectures such as convolutional neural network--long short-term memory (CNN--LSTM) models can capture spatial and temporal dynamics for continuous kinematic decoding; however, systematic residual errors persist in predicted trajectories. We propose a two-stage decoding framework that applies reinforcement learning (RL) to perform residual kinematic correction on the outputs of a CNN--LSTM decoder (CNN--LSTM--RL). The RL agent is trained offline without direct EEG input and instead operates on predicted kinematic trajectories to optimize movement accuracy relative to target trajectories. Decoding performance was quantified using Pearson correlation coefficients ($r$) and Root Mean Square Errors (RMSE) along the $x, y$, and $z$ axes. Compared to CNN--LSTM applied alone, CNN--LSTM--RL improved the mean correlation from $0.5076$ to $0.7181$ ($p = 0.0005$) in 2D and from $0.6420$ to $0.7780$ ($p = 0.0059$) in VR, with relative gains of $41.5%$ and $21.2%$, respectively. Correspondingly, RMSE was reduced from $0.0890$ to $0.0532$ (2D, $p < 0.0001$) and from $0.0714$ to $0.0441$ (VR, $p < 0.0001$), representing relative reductions of $40.2%$ and $38.2%$. These findings demonstrate that this scalable framework enhances 3D BCI MI decoding by correcting kinematic errors via offline residual RL without extra neural data, advancing neurorehabilitation, prosthetics, and virtual interaction.