Classifying Motion States from Neural Activity of Non-Human Primates for Brain-Computer Interfaces

YICONG XIAO^1*Spencer Kellis^2,3Christopher Reiche^1,4Florian Solzbacher^1,3,5,6

• ¹Department of Electrical and Computer Engineering, College of Engineering, University of Utah, Salt Lake City, Utah, United States
• ²Neurorestoration Center, Keck School of Medicine, University of Southern California, Los Angeles, California, United States
• ³Blackrock Neurotech, Salt Lake City, Utah, United States
• ⁴Department of Engineering Sciences, Jade Hochschule, Wilhelmshaven, Germany
• ⁵Department of Materials Science & Engineering, University of Utah, Salt Lake City, United States
• ⁶Department of Biomedical Engineering, University of Utah, Salt Lake City, United States

Many brain-computer interface (BCI) systems record neural activity from the sensorimotor network to drive the velocity of a mouse cursor, robotic limb, or other similar end effectors. These systems accurately translate neural activity corresponding to imagined or attempted movement into control signals for movement of an effector. However, without explicit mechanisms to recognize the intended absence of movement, decoders continue to receive nonzero inputs from the neural activity and thus infer nonzero effector displacement, even though the intended behavior of the BCI user is to maintain a constant effector state. This paper proposes a scheme to classify intended effector stationary states versus movement states directly from neural activity. In offline analysis with intracortical premotor and primary motor recordings from two non-human primates, mean classification accuracy was 0.936 and 0.930, respectively, while preserving decoded trajectory continuity. These results suggest that the proposed scheme provides a reliable and accurate means of distinguishing between stationary and movement states, offering potential benefits for stable and safe BCI control.