Complex-Valued Neural Network Phase Difference Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network technologies face challenges in accurately estimating data with significant phase correlations, as they struggle to effectively treat phase differences between neurons within the same layer, leading to deteriorated estimation accuracy, especially in applications like radar signal processing.
Innovation Solution
A neural network architecture that includes a phase difference computation layer to output phase differences between complex data sets, utilizing a first complex-valued neural network and a second complex-valued neural network, with the option to transform complex data into real-valued data for improved estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all possible combination pairs corresponding to the number of elements of the reception signal vector are input to the neural network, then the neural network can process phase difference data, but the size of the neural network tends to get larger
Solution Approach 1:
The patent extracts only the essential phase difference information from the complex data rather than processing all possible combination pairs. By computing phase differences between corresponding elements of complex data sets and using only these extracted phase difference values as input to the neural network, the system reduces the input dimensionality while preserving the critical phase correlation information needed for accurate estimation.
2Device complexity
If conventional neural networks are used to process complex data, then the network structure remains simple, but phase difference between neurons belonging to the same layer cannot be treated
Solution Approach 1:
The patent segments the complex data processing into distinct functional components: a phase difference computation unit that calculates phase differences between corresponding elements of complex data sets, and a neural network that processes these phase difference values. This segmentation allows the network to explicitly handle phase difference information while maintaining a manageable structure, as the phase difference computation is separated from the neural network processing.
3Ease of manufacture
If real-valued neurons are used in the neural network, then the network implementation is simpler, but phase correlation information is lost
Solution Approach 1:
The patent introduces phase difference values as an intermediary representation that captures the essential phase correlation information from complex data. Instead of directly processing complex data with real-valued neurons (which would lose phase information), the system computes phase differences as intermediate features and feeds these to the neural network. This intermediary approach preserves phase correlation information while allowing the use of simpler real-valued neurons for processing.
Data Source
AI summary
It is possible to improve estimation accuracy with regard to data in which significance is attached to a relative phase.Provided is an information processing device including an estimation unit configured to estimate a status by using a neural network. The neural network includes a first complex-valued neural network to which complex data is input, a phase difference computation layer from which phase difference for each element between a plurality of sets with regard to the complex data is output, and a second complex-valued neural network from which complex data is output on the basis of the phase difference.