Pixel-Sum Imaging Circuit for On-Device Neural Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current imaging devices require external processing for image data conversion and processing, leading to increased power consumption and communication latency, and lack the capability for efficient on-device data processing and intelligent functions such as image recognition.
Innovation Solution
An imaging device with a pixel block and multiple circuits that enable data retention and arithmetic processing within the device, utilizing metal oxide transistors for low off-state current and silicon transistors for high-speed operation, allowing for on-device image processing and efficient data handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image data is converted and processed externally after capture, then processing capability is improved, but communication latency and power consumption increase
Solution Approach 1:
The patent merges the imaging function with image processing functions by integrating multiple circuits (first circuit for row selection, second circuit for pixel selection and summing, third circuit for arithmetic operations) directly into the imaging device. This allows image data to be processed within the device itself rather than being transmitted externally, thereby reducing communication latency and enabling real-time processing.
Solution Approach 2:
The imaging device is designed with multi-functional circuits that can perform various operations including row selection, pixel selection, summing of pixel values, and arithmetic processing. The first circuit can select pixels from one row or multiple consecutive rows, while the second and third circuits perform different processing modes (summing mode and arithmetic operation mode), making the device universally capable of multiple image processing tasks.
2Extent of automation
If image data is converted and processed externally, then processing flexibility is improved, but power consumption increases
Solution Approach 1:
By combining imaging and processing functions in a single integrated device, the patent eliminates the need for continuous data transmission to external processors. The integrated circuits perform processing operations locally on the captured image data, significantly reducing the power consumption associated with communication and external processing while maintaining processing flexibility.
3Productivity
If multiple circuits are added for on-device processing, then processing capability is improved, but device complexity increases
Solution Approach 1:
The patent segments the image processing function into distinct modular circuits: a first circuit for row selection, a second circuit for pixel selection and summing operations, and a third circuit for arithmetic processing. This segmentation allows each circuit to perform a specific function efficiently, improving overall processing speed while making the complex functionality manageable through modular design.
Solution Approach 2:
The circuits are designed with multi-functional capabilities to reduce overall device complexity. The first circuit can select from one row or multiple consecutive rows of pixels. The second circuit can operate in summing mode or arithmetic operation mode. This universality allows fewer circuits to perform multiple tasks, thereby improving productivity without proportionally increasing device complexity.
4Adaptability or versatility
If advanced processing functions are integrated, then intelligent capabilities are improved, but manufacturing complexity increases
Solution Approach 1:
The patent divides the advanced processing functions into separate modular circuits that can be independently designed and manufactured. Each circuit (first circuit for row selection, second circuit for pixel selection and summing, third circuit for arithmetic operations) can be fabricated using standard semiconductor manufacturing processes, making the complex intelligent functions easier to manufacture through modular assembly.
Data Source
AI summary
An imaging device capable of executing image processing is provided. Analog data (image data) acquired through an imaging operation is retained in a pixel, and data obtained by multiplying the analog data by a given weight coefficient in the pixel can be extracted. The data is taken into a neural network or the like, whereby processing such as image recognition can be performed. Since an enormous amount of image data can be retained in pixels in an analog data state, processing can be performed efficiently.