Pulse-Based Imaging Circuit for Low-Noise On-Chip Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Solid-state imaging elements, such as CMOS image sensors, face issues with charge leakage and noise contamination during data readout, which affect image quality and processing efficiency.
Innovation Solution
An imaging device design that converts photoelectric conversion data into pulse signals, utilizing a pixel circuit with specific transistor and capacitor configurations to minimize noise and enable efficient image processing, including arithmetic operations and data compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data potential is held in a charge accumulation portion in a pixel, then image data can be captured, but charge leakage and noise contamination occur during readout
Solution Approach 1:
The patent extracts the data potential from the charge accumulation portion and transfers it to a floating diffusion region. This separation allows the charge accumulation portion to be isolated from readout operations, preventing charge leakage and noise contamination while maintaining the captured image data integrity.
Solution Approach 2:
The floating diffusion region acts as an intermediary between the charge accumulation portion and the readout circuitry. It receives the data potential from the charge accumulation portion and interfaces with the transfer transistor and amplifier, preventing direct interaction that would cause charge leakage and noise.
2Productivity
If image processing is performed externally, then processing can be done, but conjunction speed with external devices is reduced and peripheral device load increases
Solution Approach 1:
The patent merges image processing functions directly into the imaging device by integrating a neural network unit with the pixel array. This combination allows arithmetic operations and image recognition to be performed on-chip, eliminating the need for external data transfer and reducing processing latency.
Solution Approach 2:
The imaging device is designed with multi-functionality by incorporating both photoelectric conversion and neural network processing capabilities in a single integrated structure. This universal design enables the device to perform both image capture and complex image processing tasks without requiring separate external processing units.
3Adaptability or versatility
If more processing functions are added to the imaging device, then image processing capability is enhanced, but device complexity increases
Solution Approach 1:
The patent implements a nested structure where the neural network unit is integrated within the pixel array structure. The processing circuits are embedded alongside the photoelectric conversion elements, creating a compact hierarchical arrangement that enhances functionality while minimizing the increase in overall device complexity.
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
The solution reduces noise interference, enhances image processing speed, and lowers power consumption while maintaining high sensitivity and reliability, enabling efficient image capture and recognition.
Implementation Method 1
a first photoelectric conversion element, a first transistor, a second transistor
Data Source
AI summary
An imaging device that generates a pulse signal by utilizing photoelectric conversion operation is provided.A data potential generated by the photoelectric conversion operation is input to a pulse generation circuit to output a pulse signal having a spike waveform. In addition, a structure in which product-sum operation of pulse signals is performed is provided, and digital data is generated from a new pulse signal. The digital data is taken into a neural network or the like, whereby processing such as image recognition can be performed. Processing up to taking an enormous amount of image data into a neural network or the like can be performed in the imaging device; thus, processing can be efficiently performed.