Selective Sensing Operator for On-Sensor Data Dimensionality Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computational infrastructures, particularly in IoT applications, face challenges with data transmission and processing due to the high energy consumption of RF transceivers and the computational complexity of digital-domain compressive sensing, which limits the practical application of efficient data compression in resource-constrained devices.
Innovation Solution
A data-driven nonuniform subsampling approach called selective sensing, which co-optimizes a selective sensing operator with a subsequent information decoding neural network to reduce data dimensionality in a computation-free fashion, improving power consumption and bandwidth efficiency without requiring significant processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If digital-domain compressive sensing is implemented, then data compression ratio is improved, but computational complexity increases
Solution Approach 1:
The patent extracts only the essential information from the original signal by applying a subsampling operator that selects a small subset of measurements. This extraction approach achieves data compression without requiring complex transformation computations, as the subsampling operator simply selects predetermined portions of the input signal.
Solution Approach 2:
The patent replaces the traditional mechanical/computational compressive sensing system with a neural network-based system. The neural network learns optimal subsampling patterns and reconstruction strategies, substituting complex linear algebra computations with learned representations that require minimal computational resources during operation.
2Loss of energy
If subsampling rate is reduced, then bandwidth efficiency is improved, but information loss increases
Solution Approach 1:
The patent performs preliminary learning during the training phase where the neural network learns optimal subsampling patterns and reconstruction strategies from training data. This preliminary action enables the system to make informed decisions about which measurements to take, ensuring maximum information retention even at low subsampling rates.
Solution Approach 2:
The patent implements a feedback mechanism where the neural network uses the subsampled measurements to reconstruct the original signal and then compares the reconstruction quality. This feedback loop allows the system to iteratively improve its subsampling and reconstruction strategies, ensuring that minimal information is lost even when transmitting fewer measurements.
Data Source
AI summary
A data-driven nonuniform subsampling approach for computation-free on-sensor data dimensionality is provided, referred to herein as selective sensing. Designing an on-sensor data dimensionality reduction scheme for efficient signal sensing has long been a challenging task. Compressive sensing is a generic solution for sensing signals in a compressed format. Although compressive sensing can be directly implemented in the analog domain for specific types of signals, many application scenarios require implementation of data compression in the digital domain. However, the computational complexity involved in digital-domain compressive sensing limits its practical application, especially in resource-constrained sensor devices or high-data-rate sensor devices. Embodiments described herein provide a selective sensing framework that adopts a novel concept of data-driven nonuniform subsampling to reduce the dimensionality of acquired signals while retaining the information of interest in a computation-free fashion.