Combined
Virtual Arm, Grasp Simulator and Action Recognition Network
This integrated system is composed of a kinematics model of 19 DOF
arm, a set of grasp generation routines and a simplified version of the
mirror neuron system model published. The hand can generate three types
of grasping movements to objects at various locations. In turn mirror
neuron
model can use the 'hand state' (the relation of an hand to a target
object)
trajectory to determine the type of the grasp. Enjoy the
kinematics simulation of grasping and the response of action
recognition.
Support
Vector Machine implementations
Here you can find the classifiction and
regression
application of support vector machines with polynomial and graussian
kernels
(both with adjustable parameters). Draw a 2D function by holding
down the left mouse button and moving the mouse and click KA-SVM to
perform
the regression.
SVM regression applet: Draw
a 2D function by holding down the left mouse button and moving
the
mouse and click KA-SVM to perform the regression. You can
choose
the kernel and kernel parameters on the left panel.
SVM classification applet:
Choose your classes by clicking on Class 0 or Class 1
then draw your data points using the mouse. Click KA-SVM to
perform
the classification. Use Iteration gadget to set the epoch size for each
click. You can choose the kernel and
kernel
parameters on the left panel.
Neurobench
Neurobench is a simple web accesible tool I implemented for
visualization
and analysis of neurophysiological time series data. I used the for
visualizing
mirror neuron data. Unfortunately mirror neuron data from Rizzolatti
labs
is not publicly accesible.
Neurobench1.0 (publicly accesible, with
sample
data)
Neurobench2.0 (restricited access, due to private
data)