Edge Computing Neural Processor for Wearable Signal Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern biomedical devices, such as prosthetic devices, face challenges in designing motion classifiers due to the large number of channels and stringent communication latency requirements, leading to bottlenecks when processing sensor data in wearable devices.
Innovation Solution
The integration of an edge computing distributed neural processor with built-in machine learning capabilities and body channel communication, which reduces data traffic, power consumption, and communication latency by processing data locally and minimizing the need for centralized processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor fusion techniques are used to improve classification accuracy, then measurement precision is improved, but device complexity increases due to the large number of channels
Solution Approach 1:
The patent divides the neural network into distributed processors, each handling specific sensor channels independently. This segmentation allows the system to maintain high classification accuracy through sensor fusion while reducing the complexity burden on any single processor, as each handles only a subset of channels.
Solution Approach 2:
The patent introduces a distributed architectural dimension to handle the multi-channel data problem. Instead of processing all channels on a single processor, the system distributes the computational dimension across multiple processors, each handling specific channel data locally before aggregating results.
2Device complexity
If centralized processing is used to handle large amounts of sensor data, then device complexity is reduced, but communication latency increases and communication bandwidth is consumed
Solution Approach 1:
The patent segments the centralized processing function into multiple distributed processors, each handling local sensor data independently. This eliminates the bottleneck of all data converging to a single centralized processor, reducing communication latency and bandwidth requirements while maintaining manageable complexity at each node.
Solution Approach 2:
The patent extracts processing functionality from the centralized architecture and places it at the edge devices. By taking out computational tasks from the central processor and distributing them to edge processors, the system reduces communication overhead while maintaining overall system manageability.
3Productivity
If high-end microprocessors are used to process physiological signals, then processing capability is improved, but power consumption increases
Solution Approach 1:
The patent segments the processing task across multiple simple edge processors rather than using a single high-end microprocessor. Each edge processor handles specific sensor data with minimal computational resources, collectively achieving the required processing capability while consuming far less power than a centralized high-end processor would require.
Solution Approach 2:
The patent enables edge devices to process their own sensor data locally without requiring powerful centralized processing. Each edge processor is self-sufficient enough to handle its assigned channels independently, eliminating the need for high-power microprocessors and achieving energy-efficient operation.
Data Source
AI summary
Systems and/or methods may include an edge-computing distributed neural processor to effectively reduce the data traffic and physical wiring congestion. A local and global networking architecture may reduce traffic among multi-chips in edge computing. A mixed-signal feature extraction approach with assistance of neural network distortion recovery is also described to reduce the silicon area. High precision in signal features classification with a low bit processing circuitry may be achieved by compensating with a recursive stochastic rounding routine, and provide on-chip learning to re-classify the sensor signals.