Neural representation and identification of reaching targets by spike trains in motor cortex

Neural prostheses could help disabled patients of immobility to restore movements by exploiting brain signals. In the past decade, it has been shown great progress in intracortical neural recording techniques and neural signal processing. We review two classes of methods, population vector and maximal likelihood, for decoding the activity of primary motor cortical neurons in a nonhuman primate during center-out arm reaching movements. In particular, we show that these two methods share the same spirit of pooling activities of population of neurons but with different use of tuning function. We further compare their performance by using real neuronal data from reaching movements and inspect the effects of different parameters. It shows that maximal likelihood approach outperforms the population vector method consistently, which could be due to the more effective use of the tuning function.