Biomedical Engineering Reference
In-Depth Information
and thus provides more accurate nonlinear properties than just assuming the exponential or Gauss-
ian function. It can actually be implemented as a look up table for the nonlinear tuning function
f
in testing as
å
t
j
k k v
(
× - ×
k v
)
test
spike,training
t
j
p
(spike |
k v
× =
)
test
å
t
i
k k v
(
× - ×
k v
)
(6.19)
test
training
i
t
where
k
is the Gaussian kernel,
v
is a possible sample we generate at time
t
in the test data.
test
j
i
v
is one spike-triggered velocity vector sample in the training data, and
v
is one ve-
spike,training
training
locity vector sample in the training data.
6.4.3 Estimating the Time delay From the Motor Cortex to Movement
Before implementing the generative BMI based on spikes, one has to align the spike trains with
the corresponding hand movements, which is accomplished by estimating the causal time delays
due to propagation effects of signals in the motor and peripheral nervous system, as shown in
Figure
6.9
.
FIgURE 6.9:
Illustration of the time delay between neuron spike train (bottom plot) and the kinemat-
ics response (upper plot).
