Optimal Achievable Encoding for Brain Machine Interface

Abstract

Our goal was to develop efficient algorithms for encoding visual stimuli using an artificial retina, in a manner that optimizes the artificial visual image delivered to the brain. We did this by focusing on three components required. First, we developed novel models of retinal encoding that improve upon the state of the art, by using machine learning methods to incorporate spatial and temporal nonlinearities. Second, we developed novel methods of decoding images from retinal responses, using artificial neural networks, and by applying linear decoding to complete recorded populations of retinal ganglion cells for the first time. Third, we developed a greedy, dictionary-based encoding approach to translate a visual image into sequential patterns of electrical stimulation in real time, in a manner that optimizes visual image transfer to the brain.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 22, 2017
Accession Number
AD1047301

Entities

People

  • Eduardo Chichilnisky

Organizations

  • Stanford University

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Algorithms
  • Coding
  • Computational Neuroscience
  • Decoding
  • Dictionaries
  • Government Procurement
  • Governments
  • Images
  • Information Exchange
  • Information Processing
  • Information Systems
  • Learning
  • Machine Learning
  • Neural Networks
  • Prostheses And Implants

Fields of Study

  • Computer science

Readers

  • Computer Programming and Software Development.
  • Neural Network Machine Learning.
  • Vision Science/Vision Psychology/Cognitive Neuroscience.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks