Abstract
Matrix-vector multiplication is a fundamental operation in linear algebra. A large variety of optical processors for performing matrix multiplication have been proposed and some experimentally demonstrated. We propose the use of a pinhole hologram to perform matrix-vector multiplication. The method is capable of storing and addressing a large number of holograms inside a photorefractive crystal, and achieving better resolution while performing the multiplication. On recording, each matrix (M × N) pattern passes through a pinhole and then stores its hologram in a photorefractive crystal. A phase-conjugate beam is used to read the holograms. Among them the desired one, say matrix A, is selected by opening a corresponding pinhole. An NX1 vector B is illuminated by the selected matrix pattern. The overlap of matrix A and B produces a matrix C. Using a cylindrical lens we can convert the M × N matrix C into a MX1 vector c which is the product of A and B. A 2-D detector array is used to detect the intensity distribution of vector c and to decide its value according to some preset threshold value. This method can be programmed to implement a parallel real-time matrix-vector processor.
© 1991 Optical Society of America
PDF ArticleMore Like This
Arthur E. Chiou, Pochi Yeh, and Monte Khoshnevisan
FA1 Spatial Light Modulators and Applications (SLM) 1988
L. Scharf, W. T. Cathey, and K. M. Johnson
MD7 Optical Computing (IP) 1987
David Brady and Demetri Psaltis
MB3 OSA Annual Meeting (FIO) 1987