Faster RCNN Wireless Signal Detection in Wideband RF
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to effectively detect and localize time and frequency information of wireless signals in wideband RF spectra, limiting their ability to manage and secure wireless devices in IoT environments.
Innovation Solution
A system utilizing the Faster RCNN deep learning architecture converts RF time-series data into spectrogram images, allowing for the detection and localization of wireless signals by identifying rectangular objects representing RF transmissions, which correspond to starting time, channel frequency, and time and frequency span.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional signal detection methods are used in wideband RF spectra, then detection capability is limited, but system complexity and energy consumption increase when advanced methods are applied
Solution Approach 1:
The patent replaces conventional mechanical signal processing methods with a deep learning-based system. The Faster RCNN architecture, originally designed for image recognition, is adapted to detect and localize RF signals by treating spectrograms as images, thereby achieving superior detection accuracy without proportionally increasing system complexity
Solution Approach 2:
The patent transforms RF time-series data into spectrogram images, changing the parameter representation from temporal domain to time-frequency domain. This parameter transformation enables the application of image processing techniques to RF signal detection, improving measurement precision while leveraging existing computational frameworks
2Loss of information
If time and frequency localization is implemented for wireless signals, then spectrum management capability improves, but processing time and computational load increase
Solution Approach 1:
The patent performs preliminary transformation of RF data into spectrogram representation before detection. This pre-processing step organizes the time-frequency information in a structured format that the Faster RCNN model can efficiently process, reducing the computational burden during actual signal detection and localization
Solution Approach 2:
The patent adds a visual dimension to RF signal analysis by converting 1D time-series data into 2D spectrogram images. This dimensional transformation enables simultaneous extraction of time and frequency characteristics through spatial pattern recognition, recovering complete time-frequency information without proportional increases in processing time
3Measurement precision
If deep learning architecture is used for signal detection, then detection accuracy and localization precision improve, but energy consumption increases
Solution Approach 1:
The patent employs a universal deep learning architecture (Faster RCNN) that was originally designed for image recognition but is now applied to RF signal detection. This multi-functional approach leverages pre-trained models and shared computational components, achieving high localization precision while reducing energy consumption compared to designing and training specialized models from scratch
Data Source
AI summary
The present invention comprises a novel system and method to detect and estimate the time-frequency span of wireless signals present in a wideband RF spectrum. In preferred embodiments, the Faster RCNN deep learning architecture is used to detect the presence of wireless transmitters from the spectrogram images plotted by searching for rectangular shapes of any size, then localize the time and frequency information from the output of the FRCNN deep learning architecture.