Optical Convolution via Spatial Domain Fourier Transform
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing convolutional artificial neural networks face limitations in energy efficiency and calculation speed due to the slow processing of increasing amounts of image information, particularly in the convolution layer where most energy consumption occurs.
Innovation Solution
The method involves optically performing a convolution operation in the spatial domain by optically Fourier transforming the spatial domain kernel and image, and then performing an element-wise product operation using an opto-optical modulator, thereby eliminating the need for electronic computer processing of the Fourier transform.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If convolution operations are performed using electronic computers, then calculation accuracy is maintained, but energy consumption is high and calculation speed is slow
Solution Approach 1:
The patent replaces electronic computer-based convolution calculations with an optical computing system that uses light propagation and optical components (lenses, spatial light modulators, detectors) to perform convolution operations physically, thereby reducing energy consumption and increasing calculation speed
Solution Approach 2:
The patent transforms the computational problem into an optical domain by changing the parameter space from digital computation to optical field manipulation, using wavelength, amplitude, and phase of light to encode and process image data
2Loss of energy
If Fourier transform is performed using electronic computers, then precise mathematical transformation is achieved, but unnecessary calculations increase energy consumption
Solution Approach 1:
The patent replaces electronic Fourier transform calculations with optical Fourier transform using lens-based optical systems, where light propagation naturally performs the mathematical transformation without requiring active computational processing
Solution Approach 2:
The optical system performs Fourier transform automatically through the natural propagation of light through optical components, eliminating the need for separate computational processing steps and reducing energy consumption
3Device complexity
If spatial domain kernel is directly used in optical convolution, then device complexity is reduced, but additional Fourier transform steps are required
Solution Approach 1:
The patent implements continuous optical processing where the spatial domain kernel is optically Fourier transformed and immediately used in the convolution operation without interruption, maintaining continuous useful action throughout the processing pipeline
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach increases energy efficiency and achieves fast calculation speeds by minimizing unnecessary calculations and energy consumption typically required by electronic computers, while directly utilizing the spatial domain kernel for optical convolution operations.
Implementation Method 1
performing a first optical Fourier transform on a spatial domain image; performing a second optical Fourier transform on a spatial domain kernel
Implementation Method 2
performing an element-wise product operation between a result of the first optical Fourier transform and a result of the second optical Fourier transform; the element-wise product operation may be performed by an opto-optical modulator
Data Source
AI summary
The present disclosure relates to a method and apparatus for performing a spatial domain-based optical convolution operation. A method of performing a convolution operation according to an embodiment of the present disclosure may comprise: performing a first optical Fourier transform on a spatial domain image; performing a second optical Fourier transform on a spatial domain kernel; performing an element-wise product operation between a result of the first optical Fourier transform and a result of the second optical Fourier transform; calculating a convolution result by performing a third optical Fourier transform on a result of the element-wise product operation; and obtaining data based on the convolution result.