Neural Computer Image Sensor Photocurrent Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neural computers require separate units for image capture and feature map generation, leading to increased complexity and volume, and inefficient processing capacity and time for image classification tasks.
Innovation Solution
A neural computer design that integrates an image sensor with an optical signal processor capable of controlling photocurrent, allowing image capture and feature map generation within the same unit, using a preprocessor to generate feature maps and a flattening unit to convert them into tabular data for image classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If separate units are used for image capture and feature map generation, then functional separation is achieved, but device complexity and volume increase
Solution Approach 1:
The patent merges the image sensor and optical signal processor into a single integrated device. The image sensor captures images while the optical signal processor simultaneously generates feature maps through optical computing operations, eliminating the need for separate image capture and processing units. This integration directly reduces device complexity and volume while maintaining distinct functional capabilities.
2Productivity
If conventional neural computer architecture is used, then processing capability is achieved, but processing capacity and time are inefficient
Solution Approach 1:
The patent replaces conventional electronic signal processing with optical signal processing for feature map generation. The optical signal processor uses optical fields to perform convolution and pooling operations, which are inherently parallel and faster than sequential electronic processing. This substitution significantly improves processing capacity while reducing the time required for image classification tasks.
3Device complexity
If degree of integration is increased, then device compactness is improved, but manufacturing difficulty increases
Solution Approach 1:
The patent designs the image sensor to perform multiple functions: capturing images, generating feature maps through optical processing, and providing electrical signals for classification. This multi-functional design achieves high degree of integration while using a unified manufacturing process for the image sensor structure, avoiding the need for separate assembly of multiple discrete components and thereby reducing manufacturing difficulty.
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 integration simplifies the image processing operation, reduces processing capacity and time, and enhances the degree of integration, enabling a more compact and efficient neural computer for image classification tasks.
Implementation Method 1
an image sensor configured to receive the image and generate the feature map
Data Source
AI summary
Disclosed is a neural computer including an image sensor capable of controlling a photocurrent. The neural computer according to an embodiment includes a preprocessor configured to receive an image and generate a feature map for the received image; a flattening unit configured to transform the feature map generated by the preprocessor into tabular data to provide data output; and an image classifier configured to classify images received through the preprocessor by using the data output by the flattening unit as an input value. The preprocessor includes an optical signal processor configured to receive the image and generate the feature map.