May 2003
Volume 44, Issue 13
ARVO Annual Meeting Abstract  |   May 2003
Pattern Encoding and Data Encryption in Learning Retina Implants
Author Affiliations & Notes
  • O. Baruth
    Dept of Computer Science, University of Bonn, Bonn, Germany
  • D. Neumann
    Dept of Computer Science, University of Bonn, Bonn, Germany
  • R.E. Eckmiller
    Dept of Computer Science, University of Bonn, Bonn, Germany
  • Footnotes
    Commercial Relationships  O. Baruth, None; D. Neumann, None; R.E. Eckmiller, None.
Investigative Ophthalmology & Visual Science May 2003, Vol.44, 5051. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      O. Baruth, D. Neumann, R.E. Eckmiller; Pattern Encoding and Data Encryption in Learning Retina Implants . Invest. Ophthalmol. Vis. Sci. 2003;44(13):5051.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Abstract: : Purpose: To develop and test learning retina implant systems, which combine personalized pattern encoding with data encryption for secure wireless communication and authentication. Methods: An encryption unit (EU) as an array of 16x16 filters with selectable, pre-defined spatio-temporal properties (ST*) was developed to encode and encrypt visual patterns P1 into sets of discrete temporal patterns of defined duration. A corresponding learning decryption unit (DU) to map EU-output signals onto visual patterns P2 was developed as central visual system simulator (VS) in order to test both the retina encoder (RE) properties and the encryption properties of EU. DU was trained in conjunction with EU to decrypt any set of temporal patterns at the EU-output into corresponding visual patterns P2 with 16x16 pixels. Results: (1) The DU learning phase in conjunction with a given EU yielded high quality decryption results as indicated by Hamming distances down to 0 (range: 0-256) between DU-output P2 and corresponding EU-input P1. (2) Due to the ambiguous ST*-filter input-output properties, pattern decryption could only be achieved by evaluating the entire set of temporal EU-output patterns by means of DU. (3) ST*-filter properties could be selected to approximate typical receptive field properties of retinal neurons for the purpose of using EU as RE. (4) The specific encryption algorithm in EU as RE provided the option of authentication at an implanted receiver stage in order to assure personalized communication in learning retina implants. Conclusions: A.The authentication requirement for visual prostheses with external and implanted modules can be met by combined encoder / encryptor modules. B.The one-way function of typical spatio-temporal filters can be used for data encryption.

Keywords: pattern vision • computational modeling • retinal connections, networks, circuitry 

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.