Neural decoding based on kernel regression

作者：Yu Yi^*; Yang Nan; Dong Bingchao; Zheng Xiaoxiang

来源：International Journal of Digital Content Technology and Its Applications, 2012, 6(13): 427-435.

DOI：10.4156/jdcta.vol6.issue13.46

• Gaussian kernels
• Kernel regression
• Kernel regression model
• Motor cortex
• Neural activity
• Neural decoding
• Polynomial kernels
• Primary motor cortex
• Sigmoid kernels
• WIENER filters

This paper introduces a novel neural decoding method based on kernel regression. We estimated the performance of kernel regression model in decoding rats'; neural activity of Primary Motor Cortex during rats'; lever pressing task, and the Wiener filter was implemented here firstly as a comparison, which has been successfully applied in BMI systems. In rat S9-03, Sigmoid kernel (CC=0.8535, MSE=0.5470, p<0.001), Gaussian kernel (CC=0.8378, MSE=0.3390, p<0.001), Polynomial kernel (CC=0.8448, MSE=0.3147, p<0.001), generated more smooth and accurate outputs than Wiener filter (CC=0.7410, MSE=0.5470, p<0.001), especially in the period between pressing events. Experiments have shown that kernel regression has superior capacity for neural decoding.

• 出版日期2012