Cyclic Residual Network for Variable-Length Signal Modulation Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing modulation recognition methods are complex and inflexible, requiring manual labor and being unable to process signals of variable lengths due to the use of deep residual networks that can only handle fixed-length inputs.
Innovation Solution
A method and apparatus using a cyclic residual network that processes signals by extracting real and imaginary parts, converting them into amplitude-and-phase feature matrices, and inputting these into a pre-trained network with gated recurrent units (GRUs) to recognize modulation modes, allowing for flexible recognition of signals of any length.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep residual network model is used for modulation recognition, then recognition accuracy is improved, but the system cannot process signals of variable length and requires fixed-length input only
Solution Approach 1:
The patent changes the architectural parameters of the neural network by introducing cyclic residual connections and GRU units, transforming the fixed-length deep residual network into a variable-length capable cyclic residual network while maintaining recognition accuracy
Solution Approach 2:
The patent makes the network structure dynamic by using GRU units that can adaptively process sequences of varying lengths, allowing the network to handle both fixed and variable length inputs effectively
2Measurement precision
If time-frequency analysis and gray-scale image conversion are used for signal processing, then modulation recognition is achieved, but the implementation complexity increases and manual labor is required
Solution Approach 1:
The patent extracts only the essential real and imaginary parts of the signal directly, eliminating the unnecessary intermediate steps of time-frequency analysis and gray-scale image conversion, thereby simplifying the processing pipeline
Solution Approach 2:
The patent replaces the mechanical/manual process of time-frequency analysis and image conversion with direct mathematical extraction of signal components, automating and simplifying the process
3Ease of manufacture
If fixed-length input requirement is imposed on the network, then training simplicity is maintained, but the recognition method becomes less flexible for practical applications
Solution Approach 1:
The patent creates a universal network architecture that can handle both fixed and variable length inputs through cyclic residual connections and GRU units, making the model applicable to diverse practical scenarios without sacrificing training simplicity
Data Source
AI summary
The embodiments of the present application provide a method and apparatus for modulation recognition of signals based on cyclic residual network, the method comprises: obtaining a signal matrix of a to-be-recognized signal, and extracting real part information and imaginary part information of the signal matrix; generating, according to extracted real part information and imaginary part information, a real-and-imaginary-part feature matrix of the to-be-recognized signal; converting, according to a preset matrix conversion method, the real-and-imaginary-part feature matrix into an amplitude-and-phase feature matrix; and inputting the amplitude-and-phase feature matrix into a pre-trained cyclic residual network to obtain a modulation mode corresponding to the to-be-recognized signal. In the embodiments of the present application, the processing of the to-be-recognized signal is simple and easy to operate, in which neither complex algorithms nor manual processing is required, the flexibility of recognition is high, and the result of modulation recognition of the to-be-recognized signal can be accurately obtained.