AI Sensor Data Processing Circuit with Segmented Neural Network Acceleration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Real-time response processors in electronic devices are limited to handling simple tasks due to high energy consumption and low identification capabilities, making it difficult to process complex scenarios without excessive power usage.
Innovation Solution
An integrated circuit with a real-time response processor and an accelerator that uses a simplified neural network model to identify complex scenarios, allowing for real-time processing of complex tasks without significant power consumption increases, by extracting target data and determining operations based on identification results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a real-time response processor operates continuously to respond to external sensor data, then real-time response capability is improved, but energy consumption increases excessively
Solution Approach 1:
The system segments processing tasks between two processors: a first processor that continuously monitors sensor data with low power consumption, and a second processor that performs complex identification tasks only when needed. This segmentation allows real-time monitoring while avoiding continuous high energy consumption.
Solution Approach 2:
The first processor periodically checks sensor data and only triggers the second processor when specific conditions are met (e.g., detecting a wake-up word or specific pattern). This periodic action pattern reduces energy consumption compared to continuous operation of the second processor.
2Adaptability or versatility
If the real-time response processor handles complex tasks, then task processing capability is improved, but energy consumption increases excessively
Solution Approach 1:
Complex identification tasks are segmented and assigned to a dedicated second processor (accelerator) that specializes in neural network processing. The first processor handles simple monitoring tasks. This segmentation allows complex task processing without requiring the main processor to continuously operate at high capacity.
Solution Approach 2:
The first processor acts as an intermediary that filters sensor data and only transmits relevant data to the second processor when identification is needed. This intermediary role reduces the energy burden on any single processor while maintaining complex task processing capability.
3Speed
If a simplified neural network model is used in the accelerator, then processing speed is improved, but identification accuracy may deteriorate
Solution Approach 1:
The system changes parameters of the neural network model (such as reducing layers, neurons, or complexity) to create a simplified version that runs faster on the accelerator. This parameter adjustment balances processing speed with acceptable identification accuracy for the specific application requirements.
Data Source
AI summary
This disclosure discloses an integrated circuit in the artificial intelligence field. This disclosure provides an integrated circuit. The integrated circuit includes a first processor, configured to obtain first sensor data from a first external sensor, and extract first target data from the first sensor data, where the first processor is a real-time response processor; and an accelerator, configured to identify the first target data based on a first neural network model to obtain a first identification result, where the first identification result is used to determine a target operation corresponding to the first identification result. This disclosure provides an integrated circuit and a sensor data processing method, to enable a real-time response processor to identify a complex scenario and process a complex task when responding to an external sensor in real time.