Deep Learning Spread Code Design for Active Terminal Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In massive machine-type communication environments, existing communication technologies face challenges in efficiently detecting active terminals due to high device density and resource limitations, leading to increased cross-correlation values and reduced detection performance.
Innovation Solution
An end-to-end deep neural network is employed to learn and generate spread codes suitable for various communication environments, reducing cross-correlation values and improving active terminal detection by using a first learning network to transmit signals and a second learning network to determine terminal activity, incorporating techniques like batch normalization and residual networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional spread codes with minimized cross-correlation values are used, then code design is simplified, but active terminal detection performance deteriorates in massive machine-type communication environments with different activation frequencies
Solution Approach 1:
The patent changes the fundamental parameter of spread code design from minimizing cross-correlation values to maximizing detection accuracy through deep learning optimization. The spread codes are no longer designed using traditional mathematical criteria but are instead learned from data to adapt to specific communication environments, particularly environments with varying device activation frequencies.
Solution Approach 2:
The patent replaces traditional mathematical algorithms for spread code optimization with deep learning-based artificial neural networks. Instead of using deterministic mathematical methods to minimize cross-correlation, the system uses data-driven learning to automatically discover optimal spread code characteristics for massive machine-type communication scenarios.
2Measurement precision
If deep learning techniques are applied to learn spread codes, then active terminal detection performance improves, but system complexity increases
Solution Approach 1:
The deep learning model performs self-service by automatically learning optimal spread codes and detection strategies from communication data without requiring manual configuration or optimization. The system trains neural networks to autonomously adapt to different communication environments, eliminating the need for complex manual tuning while achieving superior detection performance.
3Loss of time
If grant-free non-orthogonal multiple access is used, then resource allocation overhead is reduced, but active terminal detection becomes more challenging due to non-orthogonal signals
Solution Approach 1:
The patent implements feedback mechanisms where the deep learning model continuously learns from detection outcomes and communication patterns. The system uses detected terminal activities and signal characteristics to refine spread code selections and detection strategies, creating a closed-loop system that adapts to the challenges of non-orthogonal multiple access while maintaining low overhead.
Data Source
AI summary
Embodiments of the present disclosure provide an active terminal detection method and an active terminal detection device that increase the performance of determining whether a terminal is active by designing a spread code to reduce a cross-correlation value of the spread code of a terminal with a high activation frequency in a massive machine-type communication environment by using deep learning.