Identification of Learning in Visuomotor
Tasks |
The brain is composed of two broad classes of cells: neurons and glia. Neurons connect with thousands of other neurons to form large networks. Specialized networks make up systems that are the basis of perception, different types of action, and high cognitive function. We are interested in understanding the function and structure of the network used to learn a motor task. If we had a better understanding of how the brain assimilates information, many beneficial applications could be developed to assist individuals to learn at an accelerated pace, test for trainability, improve safety detection, and much more.
The aim of this research is to use control system techniques to model a compensatory tracking task that has a Uniform White Noise (UWN) input with frequencies of 10Hz or lower. Participants are tasked to center a cursor on a computer screen using an isometric force task via pronation-supination movement. The task is meant to be difficult because we want to quantify the participant's performance bandwidth. System identification frequency domain analysis is used to quantify participant performance and pre-process input/output data for utilization in both polynomial and subspace identification models. Parameters in both models are related with each other and both models are validated against a validation data set.
Subspace identification models are superior to polynomial models in terms of describing the compensatory tracking task input/output data. However, there is still much room for improvement. In order to improve the subspace model output estimates and uncover model structure, we propose to use fMRI data as input and output measures to the model. In addition, optimal learning analysis will be performed on both subspace models to understand how participants learn over time and what mechanistic trade-offs are made.
 |
| Figure 1: The bottom right picture is of the experimental set-up. Each participant performs the isometric force task using pronation-supination movement. The top left picture is what each participant will see on the screen. The task is to center the red cursor using the knob. Each trial consists of a 5 second rest and a 10 second run. |
|
|