Spectrum Prediction Using N-Order Markov Channel Correlations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current spectrum prediction methods using first-order Markov models have low prediction accuracy due to neglecting historical channel state correlations and assuming independent channels, leading to high miss and omission ratios and unreliable spectrum hole selection.
Innovation Solution
A method and apparatus for spectrum prediction that involves obtaining and analyzing sampling data to generate matrices representing channel state information, extracting spectrum resource occupancy modes, and predicting future channel states by matching current data against historical patterns, thereby accounting for channel correlations and improving prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the first order Markov model is used for spectrum prediction, then the prediction process is simple, but the prediction accuracy is low due to neglecting historical channel state correlations
Solution Approach 1:
The patent transitions from a first-order Markov model that only considers the immediate previous state to an n-order Markov model that incorporates historical channel state information from multiple previous timeslots. This dimensional expansion in the historical time dimension allows the system to capture long-term channel state correlations while maintaining a structured prediction framework, thereby improving prediction accuracy without excessive complexity increase.
Solution Approach 2:
The patent performs preliminary extraction of channel state information from historical sampling data and pre-establishes the n-order Markov model with transition probability matrices before actual spectrum prediction. By preparing the historical data structure and model parameters in advance, the system reduces real-time computational complexity while ensuring accurate prediction when needed.
2Device complexity
If channels are assumed as independent from each other, then the analysis process is simplified, but the correlativity between channels is not considered leading to high miss ratio and omission ratio
Solution Approach 1:
The patent merges the state information of multiple channels by constructing a joint channel state vector that combines individual channel states into a unified prediction model. This merging approach captures the correlativity between channels while maintaining a systematic analysis framework, improving spectrum hole selection reliability without prohibitively increasing computational complexity.
Solution Approach 2:
The patent segments the spectrum into multiple channels and analyzes their individual states while simultaneously considering their correlations through the n-order Markov framework. By segmenting the analysis into manageable channel-state components that can be independently observed but jointly predicted, the system balances analytical simplicity with accurate correlation modeling.
3Speed
If the sensing time is kept short for real-time spectrum detection, then the response speed is improved, but the spectrum sensing technique has difficulty in accurately detecting idle bands
Solution Approach 1:
The patent performs preliminary spectrum sensing and collects channel state information in advance, storing this data for subsequent prediction analysis. By conducting initial sensing operations and preparing historical data beforehand, the system enables faster real-time decision-making without sacrificing detection accuracy, as the heavy sensing burden is distributed over time rather than concentrated in a single short interval.
Solution Approach 2:
The patent implements a feedback mechanism where historical spectrum sensing results are continuously fed into the n-order Markov prediction model. This feedback loop allows the system to refine its predictions based on accumulated historical data, improving detection accuracy over time while maintaining fast response speeds for real-time spectrum access decisions.
Data Source
Figure 1~3
Figure 4~6
Figure 7~8
AI summary
Method, system and apparatus for spectrum prediction are provided in the embodiments of the present invention. The method includes: obtaining the first sampling data of a target spectrum, wherein the first sampling data including the existing service information, channel information and channel state information of the target spectrum; extracting the channel state information of all channels of the same service in each timeslot from the first sampling data, and generating a sampling matrix; extracting a spectrum resource occupancy mode combination of the target spectrum from the sampling matrix; obtaining the second sampling data of the target spectrum, matching the spectrum resource occupancy mode combination according to the second sampling data of the target spectrum, and predicting the channel state of the target spectrum in a future timeslot according to the matching result. With the embodiments of the present invention, the spectrum prediction is performed according to the correlativity between different channels of the same service, which reduces the miss ratio and omission ratio in the spectrum prediction process, and improves the spectrum prediction accuracy.