Wearable Device Deep Learning Accelerator Reduces Energy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wearable electronic devices face challenges in efficiently processing sensor data with existing technologies, leading to high energy consumption and prolonged computation times, particularly when implementing artificial neural networks (ANNs) for intelligent monitoring applications.
Innovation Solution
Incorporating a deep learning accelerator (DLA) with random access memory (RAM) in wearable devices, which includes specialized hardware for parallel vector and matrix calculations, and neuromorphic memory for reduced power consumption, enabling autonomous processing of sensor data without reliance on external processors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing technologies are used for processing sensor data in wearable devices, then the device can perform basic monitoring functions, but energy consumption is high and computation time is prolonged
Solution Approach 1:
The patent segments the computational workload by introducing a dedicated deep learning accelerator (DLA) module that handles neural network computations separately from the main processor. This segmentation allows the main processor to remain in low-power states while the DLA efficiently processes sensor data using specialized hardware architectures optimized for parallel matrix and vector operations inherent in deep learning models.
Solution Approach 2:
The patent introduces an intermediary deep learning accelerator module that sits between the sensor data acquisition system and the main processor. This DLA acts as a mediator that pre-processes and filters sensor data using lightweight neural networks before transmitting relevant information to the main processor, thereby reducing the computational burden and energy consumption of the main system while maintaining high processing speed for critical functions.
2Loss of time
If existing technologies are used for processing sensor data in wearable devices, then basic monitoring can be performed, but computation time is prolonged
Solution Approach 1:
The patent replaces the traditional general-purpose processor architecture with a specialized deep learning accelerator that uses hardware-oriented neural network computations. This substitution introduces dedicated matrix multiplication units, vector processing units, and activation function blocks that directly implement neural network operations in hardware, dramatically reducing computation time compared to software-based processing on conventional processors.
Solution Approach 2:
The patent implements continuous processing pipelines within the deep learning accelerator that maintain uninterrupted data flow from sensor input through multiple neural network layers to output generation. The architecture employs parallel processing channels and pipelined operations that eliminate idle time between computational stages, ensuring continuous useful action and minimizing overall computation time for real-time monitoring applications.
3Reliability
If all sensor data is transmitted and processed externally, then comprehensive analysis can be achieved, but data transmission requirements and energy consumption increase
Solution Approach 1:
The patent extracts and processes critical features directly within the wearable device using embedded deep learning models. The DLA identifies and extracts relevant patterns, anomalies, and key metrics from sensor data locally, transmitting only these extracted features or alerts to external systems rather than transmitting all raw sensor data. This extraction approach maintains monitoring accuracy for critical events while dramatically reducing data transmission requirements and associated energy consumption.
Solution Approach 2:
The patent enables the wearable device to perform self-service intelligent monitoring by incorporating on-device deep learning capabilities. The embedded neural networks autonomously analyze sensor data, detect anomalies, generate alerts, and make decisions without requiring constant external processing. This self-service approach ensures reliable local monitoring while minimizing energy loss from continuous data transmission to external systems.
Data Source
AI summary
Systems, devices, and methods related to a deep learning accelerator and memory are described. For example, a wearable electronic device may be configured to execute instructions with matrix operands and configured with: a housing to be worn on a person; a sensor having one or more sensor elements generate measurements associated with the person; random access memory to store instructions executable by the deep learning accelerator and store matrices of an artificial neural network; a transceiver; and a controller to monitor an output of the artificial neural network, generated using the measurements as an input to the artificial neural network. Based on the output, the controller may control selective storage of measurement data from the sensor, and/or selective communication of data from the wearable electronic device to a separate computer system.