My name is Sasha and I am passionate about the complexity of human hands. With 27 degrees-of-freedom, controlled by 17 muscles, hands are an intricate biological system, which makes replacing or even improving their function a challenging yet exciting task. Over the course of eight years conducting academic research, I have been working on a multitude of projects to assist or replicate the functionality of a human hand. During my undergraduate years, I worked with Dr. Kat Steele on developing an open-source 3D-printed wrist-driven orthosis for individuals with cervical spinal cord injury. Then, as a PhD student under the supervision of Dr. Sandro Mussa-Ivaldi and Dr. Eric Rombokas, I attempted to simplify the complexity of controlling myoelectric hand prostheses via machine learning algorithms. In my current postdoctoral project, I am working on understanding how inexpensive hand-tracking systems can be utilized for therapeutic exercises to improve hand function in a variety of populations. My ultimate goal is to continue research in an academic setting, focusing on the complexity of human hands in the field of rehab and assistive technologies.
Learning to operate a high-dimensional hand via a low-dimensional controller
In this paper, we explore how various training practices can aid learning of an autoencoder-based nonlinear controller. The controller allows an individual to control high-dimensional systems such as a 17D virtual hand via low-dimensional 2D space. We found that what truly aided the user’s ability to learn the controller was the training that established an explicit connection between the low-dimensional control space and the high-dimensional hand movements.