MIVIP: Microsystem based visual prosthesis

• Research unit: EU RFP
• Project number: 96.0156

Key words (English)

Artificial retina; visual prosthesis; micro-electronics

Alternative project number (English)

EU project number: EP 22.527

Research programs (English)

EU-programme: 4. Frame Research Programme - 1.3 Telematic systems

Short description (English)

See abstract

Partners and International Organizations (English)

Laboratoire de Microélectronique et le Laboratoire de Génie de Réhabilitation Neuronale de l'Université Catholique de Louvain (B), CNM Centro Nacional de Microelectronica (E), Matra Electronique SA (F), Neurotech SA (B), Case Western Reserve University (USA)

Abstract (English)

The aim of the project was to develop and implement a miniature camera as a front-end for a visual prosthesis. The camera is to be placed on the spectacle of a blind patient and work as an artificial eye. The final product is a 2´4´1 cm3 board featuring a miniature wide-angle optics, a CMOS retina ASIC, and a few RS-466 transceivers. It will be connected to an external processor carried by the patient through two 1.5 mm Æ and 1.5 m long cables. This link carries data from and to the ASIC using a 6 Mbits/s half-duplex serial protocol specially developed for this application.
The core of the camera is a CMOS retina ASIC featuring: 1) exposure-time-independent current mode imaging, 2) local adaptation to illumination with tuneable adaptation range, 3) A/D conversion of the image with normalisation to the brightest pixel, 4) irregular log-polar-like resolution organised as a 12 ´ 12 pixels fovea surrounded by a 15 rings ´ 48 radii periphery, 5) pixel-level large dynamic range velocity measurement at the 8 peripheral rings, 6) digital exponential encoding of measured velocity, 7) random selection of image acquisition ring, random access to its individual pixel after acquisition, fully random access to velocity pixels, 8) on-chip digitally programmable analogue biasing and 9) half-duplex serial data transfer peripheral. The original level adaptation algorithm proved to work satisfactorily. It is of particular interest for the specified 120° field of view within which bright and dark regions are likely to coexist. However level adaptation is not well exploited on the reduced number of pixel imposed by the log-polar-like resolution and the low fill-factor needed to implement pixel-level image processing. Velocity measurement is based on an original pixel-level exponential timer that counts at a rate inversely proportional to its current value. Using it to measure the time for an image feature (edge appearance) to travel between two neighbouring pixels allows velocities spanning more than two decades to be encoded within 5 bits while ensuring a constant relative error of about 25%.
A miniature 2-lenses objective (3 mm Æ 3 mm long) was designed. Its 120° field of view defocuses the image periphery, which is needed to suppress spatial aliasing due to the very low fill-factor at the periphery. Its relative aperture is F/2.2. Despite its delicate manufacturing, this optics fulfils the project requirements.
Together with the miniature camera, a flexible test-bench was developed to emulate the external processor developed by project partners. It features a CPLD implementing the master peripheral of the half-duplex serial transmission, which could advantageously be implemented beside the processor.

References in databases (English)

Swiss Database: Euro-DB of the
State Secretariat for Education and Research
Hallwylstrasse 4
CH-3003 Berne, Switzerland
Tel. [phone]
Swiss Project-Number: 96.0156