A learning-based approach to artificial proprioception
Event details
| Date | 02.02.2015 |
| Hour | 14:00 › 15:00 |
| Speaker | Prof. Philip Sabes, University of California San Francisco, USA. |
| Location | |
| Category | Conferences - Seminars |
Proprioception—the sense of the body's position in space—is important to natural movement planning and execution and will likewise be necessary for successful motor prostheses and brain–machine interfaces (BMIs). I will present our recent work on the development of a learning-based approach to delivering artificial proprioceptive feedback.
This work is motivated by the theoretical observation that movement planning and control rely on information from multiple sensory modalities, and that these signals are combined in a statistically optimal and highly adaptive manner. We have shown how a simple network model can learn to perform such multisensory processing, driven only by the common statistics of its inputs, e.g., by spatiotemporal correlations between sensory modalities. When then demonstrated that the same principle can be used to train animals to use an artificial sensory signal. In particular, we paired known visual feedback with an initially unfamiliar (and non-biomimetic) multichannel intracortical microstimulation signal that provided continuous information about hand position relative to an unseen target. After learning, the animals were able to use this signal to guide naturalistic movements. Furthermore, they combined the artificial signal with vision to form an optimal estimate of hand position. These results demonstrate that a learning-based approach can be used to provide a rich artificial sensory feedback signal, suggesting a new strategy for restoring proprioception to patients using BMIs, as well as a powerful new tool for studying the adaptive mechanisms of sensory integration.
Bio: Dr. Philip Sabes is Professor of Physiology at the University of California, San Francisco. He is also the director of the UCSF Swartz Center for Theoretical Neurobiology. Dr. Sabes' laboratory works toward understanding how the brain controls movement, and in particular the role of sensory information and learning. His lab is also using this understanding to develop Brain Machine Interfaces to help people with severe sensory and motor loss, such as spinal cord injury. Dr. Sabes was a Marshall Scholar before earning his PhD in Brain and Cognitive Sciences at MIT. He was a Sloan Research Fellow and a McKnight Scholar. He and two members of his lab were awarded the 2013 Annual BCI (Brain Computer Interface) Research Award for their work on the development of artificial somatosensory feedback. He currently holds the Jack D. and DeLoris Lange Endowed Chair in Cell Physiology.
This work is motivated by the theoretical observation that movement planning and control rely on information from multiple sensory modalities, and that these signals are combined in a statistically optimal and highly adaptive manner. We have shown how a simple network model can learn to perform such multisensory processing, driven only by the common statistics of its inputs, e.g., by spatiotemporal correlations between sensory modalities. When then demonstrated that the same principle can be used to train animals to use an artificial sensory signal. In particular, we paired known visual feedback with an initially unfamiliar (and non-biomimetic) multichannel intracortical microstimulation signal that provided continuous information about hand position relative to an unseen target. After learning, the animals were able to use this signal to guide naturalistic movements. Furthermore, they combined the artificial signal with vision to form an optimal estimate of hand position. These results demonstrate that a learning-based approach can be used to provide a rich artificial sensory feedback signal, suggesting a new strategy for restoring proprioception to patients using BMIs, as well as a powerful new tool for studying the adaptive mechanisms of sensory integration.
Bio: Dr. Philip Sabes is Professor of Physiology at the University of California, San Francisco. He is also the director of the UCSF Swartz Center for Theoretical Neurobiology. Dr. Sabes' laboratory works toward understanding how the brain controls movement, and in particular the role of sensory information and learning. His lab is also using this understanding to develop Brain Machine Interfaces to help people with severe sensory and motor loss, such as spinal cord injury. Dr. Sabes was a Marshall Scholar before earning his PhD in Brain and Cognitive Sciences at MIT. He was a Sloan Research Fellow and a McKnight Scholar. He and two members of his lab were awarded the 2013 Annual BCI (Brain Computer Interface) Research Award for their work on the development of artificial somatosensory feedback. He currently holds the Jack D. and DeLoris Lange Endowed Chair in Cell Physiology.
Links
Practical information
- Informed public
- Free
Organizer
- Prof José del R. Millán
Contact
- Christel Daidié