# Classifying motion states from neural activity of non-human primates for brain-computer interfaces

## TL;DR

This paper introduces a new method to distinguish between movement and stationary states in brain-computer interfaces using neural activity from non-human primates, improving control stability.

## Contribution

A novel neural-state classification framework (cpSVM) that uses correlation-informed dimensionality reduction and SVM classification to improve BCI control.

## Key findings

- The cpSVM achieved mean classification accuracies of 0.936 and 0.930 across two non-human primates.
- The cpSVM outperformed conventional threshold-crossing methods in accuracy, sensitivity, and specificity.
- The framework reduced spurious state transitions and improved output continuity in BCI control.

## Abstract

Brain-computer interface (BCI) systems commonly decode neural activity from sensorimotor areas to generate continuous control signals for cursors, robotic limbs, or other effectors. Although these decoders perform well during intended movement, neural activity persists during periods of intended non-movement, which can lead to unintended effector activation and reduced control stability. Accurately identifying intended stationary states therefore represents a key component for achieving stable and reliable BCI control.

We propose a neural-state classification framework (cpSVM) that distinguishes stationary and movement states directly from intracortical neural activity. This model combines principal component analysis, correlation-based feature selection, and a linear support vector machine classifier. Offline evaluations were performed using multi-unit recordings from the premotor and primary motor cortices of two non-human primates during a center-out cursor task. Performance was compared against a conventional kinematics-based threshold-crossing method.

Correlation-informed dimensionality reduction revealed a clear low-dimensional separation between stationary and movement states, supporting the selection of task-relevant neural features. The cpSVM achieved high classification performance, with mean accuracies of 0.936 and 0.930 across the two subjects. Compared with the threshold-crossing method, the cpSVM consistently improved accuracy, sensitivity, specificity, and F-score, while substantially reducing spurious state transitions and improving output continuity.

These findings demonstrate that stationary and movement states can be reliably distinguished from intracortical neural signals using a low-dimensional, correlation-informed classification approach. The proposed framework provides a promising strategy to suppress unintended effector activation and improve continuity and stability in BCI control systems.

The flow chart of the new model (cpSVM). Dash frame is the flow chart of the cPCA. The figure is divided into a training section (top green box) and a test section (bottom yellow box). In the training section, the blue dashed frame highlights the cPCA procedure used to select task-relevant principal components (PCs). The blue boxes and blue arrows represent the PCA and correlation steps that identify these PCs. Black arrows indicate the process in which the selected PC projections and labels are used to train the SVM classifier. The purple dashed arrows illustrate the dependencies between the training and test sections: (1) test data are projected onto the PC subspace selected during training, and (2) the trained SVM model is applied to these projections to generate predictions. The red arrows depict the complete test pipeline.Flowchart illustrating a machine learning process split into a training section and a test section. In the training section, training data are processed through principal component analysis (PCA), projected onto principal component (PC) subspaces, correlated, and selected PCs used to create projections and train a support vector machine (SVM). In the test section, test data are projected using selected PCs and evaluated by the trained SVM to produce output.

The flow chart of the new model (cpSVM). Dash frame is the flow chart of the cPCA. The figure is divided into a training section (top green box) and a test section (bottom yellow box). In the training section, the blue dashed frame highlights the cPCA procedure used to select task-relevant principal components (PCs). The blue boxes and blue arrows represent the PCA and correlation steps that identify these PCs. Black arrows indicate the process in which the selected PC projections and labels are used to train the SVM classifier. The purple dashed arrows illustrate the dependencies between the training and test sections: (1) test data are projected onto the PC subspace selected during training, and (2) the trained SVM model is applied to these projections to generate predictions. The red arrows depict the complete test pipeline.