Biomedical Engineering Reference
In-Depth Information
FIgURE 6.5:
Block diagram of the linear-nonlinear-Poisson model.
For the BMI, the velocity of the hand movement is measured during a [-300, 500]-msec
window with respect to every spike. Because the task is dynamic, a velocity vector at each time step
is considered as the input (equal to the size of the space for simplicity). The linear filter projects the
velocity vector
v
into its weight vector
k
(representing a direction in space), which produces a scalar
value that is converted by a nonlinear function
f
and applied to the Poisson spike-generating model as
the instantaneous conditional firing probability for that particular direction in space
p
(spike |
k v
. The
×
)
T
-
1
filter weights are obtained optimally by least squares
k
, where
E
v
|spike
is the conditional expectation of the velocity data given the spikes, which corresponds to the cross-
=
(
E v
[
v
]
+
a
I
)
E
[ ]
v
v spike
|
neuron 72: VpS PCA
25
Vp
VpS
20
15
10
5
0
-5
-10
-15
-20
-25
-30
-20
-10
0
10
20
30
1st Principal Component
FIgURE 6.6:
PCA on spike-triggered velocity vectors compared with the projection of all velocity
vectors.
