Adaptive Algorithms for Point Processes - Brain-Machine Interface Engineering - page 195

trains. We investigated an information theoretical tuning depth to evaluate the neuron tuning prop-
erties. To describe the functional relationship between neuron firing and movement, a parametric
LNP model was utilized. With the knowledge gained from the neuron physiology tuning analysis,
a novel signal processing algorithm based on a Monte-Carlo sequential estimation could be applied
directly to point processes to convert the decoding role of a brain-machine interface into a prob-
lem of state sequential estimation. The Monte Carlo sequential estimation modifies the amplitude
of the observed discrete neural spiking events by the probabilistic measurement contained in the
posterior density. The Monte Carlo sequential estimation provided a better approximation of the
posterior density than point process adaptive filtering with a Gaussian assumption. Compared with
the Kalman filter applied to a cursor control task, the preliminary kinematics reconstruction using
the Monte Carlo sequential estimation framework showed better correlation coefficients between
the desired and estimated trajectory.
However, the signal processing methodologies of spike train modeling are not yet equipped
to handle the stochastic and point process nature of the spike occurrence. Many parameters are as-
sumed and need to be estimated with significant complexity. Conventional signal processing meth-
ods on random processes work easily with optimal projections; therefore, it would be very useful
to create directly from spike trains a metric space that would allow inner product computations
because all the available tools of binned methods could be applied immediately.
REFERENCES
1. Simoncelli, E.P., et al., Characterization of neural responses with stochastic stimuli . 3rd ed. The
New Cognitive Neuroscience. 2004, Cambridge, MA: MIT Press.
2. Paninski, L., et al., Superlinear population encoding of dynamic hand trajectory in primary motor cortex .
Journal of Neuroscience, 2004. 24 (39): pp. 8551-8561. doi:10.1523/JNEUROSCI.0919-04.2004
3. Georgopoulos, A.P., A.B. Schwartz, and R.E. Kettner, Neuronal population coding of movement
direction. Science, 1986. 233 (4771): pp. 1416-1419.
4. Schwartz, A.B., D.M. Taylor, and S.I.H. Tillery, Extraction algorithms for cortical control of arm
prosthetics. Current Opinion in Neurobiology, 2001. 11 (6): pp. 701-708. doi:10.1016/S0959-
4388(01)00272-0
5. Brown, G.D., S. Yamada, and T.J. Sejnowski, Independent component analysis at the neural cocktail
party. Trends in Neurosciences, 2001. 24 (1): pp. 54-63. doi:10.1016/S0166-2236(00)01683-0
6. Eden, U.T., et al., Dynamic analysis of neural encoding by point process adaptive filtering. . Neural
Computation, 2004. 16 : pp. 971-998. doi:10.1162/089976604773135069
7. Wang, Y., A.R.C. Paiva, and J.C. Principe. A Monte Carlo sequential estimation for point process
optimum filtering, , in IJCNN. 2006. Vancouver.