Channel Prediction Using Vocabulary of Channel Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing channel estimation algorithms in wireless communications systems have low accuracy due to the complex nature of wireless channels, which affects the prediction of channel coefficients and leads to inefficiencies in data transmission.
Innovation Solution
A channel prediction method that uses a vocabulary of channel changes to predict complex-valued channel coefficients by converting change values into indices and using a neural network model for improved accuracy, retaining both amplitude and phase information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing channel estimation algorithms are used, then the channel coefficient can be estimated, but the accuracy of the channel coefficient estimation is low
Solution Approach 1:
The patent segments the channel coefficient prediction problem into two separate prediction tasks: amplitude prediction and phase prediction. By dividing the complex-valued channel coefficient into its real and imaginary components (or magnitude and phase), each component can be predicted independently using separate neural networks, thereby improving overall prediction accuracy compared to treating them as a single complex value.
Solution Approach 2:
The patent transforms the complex-valued channel coefficient prediction problem into a multi-dimensional real-valued prediction problem. Instead of predicting a single complex value, the system predicts multiple real-valued parameters (amplitude and phase separately, or real and imaginary parts separately), adding dimensional decomposition to enhance prediction precision while maintaining the ability to reconstruct the original complex channel coefficient.
2Productivity
If a neural network with real-valued weights is used for channel prediction, then future channel state information can be forecasted, but the original information of the channel coefficient (amplitude and phase) is not fully retained
Solution Approach 1:
The patent applies segmentation by separating the channel coefficient information into distinct components (amplitude and phase, or real and imaginary parts) and using separate neural networks for each component. This ensures that the original information structure is preserved while enabling effective prediction through real-valued neural networks.
Solution Approach 2:
The patent changes the parameter representation from complex-valued to separate real-valued parameters (amplitude and phase). By transforming the complex channel coefficient into its constituent real parameters, the system can utilize real-valued neural networks while maintaining complete information about the original channel coefficient through proper parameter reconstruction.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Embodiments of this application disclose a channel prediction method and a related device. The method includes: obtaining a first channel coefficient sequence in a first time period, where the first channel coefficient sequence includes a plurality of complex values of a channel coefficient; and determining a prediction value of the channel coefficient in a second time period based on the first channel coefficient sequence and a preset vocabulary of channel changes, where the vocabulary of channel changes includes a mapping relationship between a channel change value index and each change value of the channel coefficient, and the second time period is later than the first time period. According to the embodiments of this application, channel prediction accuracy can be improved.