Neural Iterative Demapping With Decoder Feedback for BER Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current communication systems face challenges in accurately estimating data transmitted by a transmitter at a receiver, particularly due to sub-optimal bit labelling schemes leading to reduced bit error rate (BER) performance.
Innovation Solution
Implementing a receiver algorithm with trainable weights that operates on channel outputs and feedback from an outer channel decoder, using iterative demapping and decoding techniques, and incorporating a neural network to generate refined estimates of transmitted data based on received data and error correction algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional bit labelling schemes are used in communication systems, then the system structure remains simple, but the bit error rate performance deteriorates
Solution Approach 1:
The patent applies dynamics by making the receiver algorithm adaptive through trainable weights that are optimized during a training phase. The receiver algorithm dynamically adjusts its parameters based on the specific bit labelling scheme being used, allowing it to adapt to different modulation schemes and channel conditions while maintaining good BER performance without requiring complex redesign for each scenario
Solution Approach 2:
The patent implements feedback through an iterative detection process where the receiver algorithm generates initial estimates, passes them through a channel decoder, and uses the decoded information as feedback to refine subsequent estimates. This feedback loop continues until convergence, significantly improving BER performance by continuously refining the detection based on decoded feedback information
2Measurement precision
If iterative detection with feedback is implemented, then data estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent applies partial action by allowing the iterative detection process to stop when a predefined maximum number of iterations is reached or when convergence criteria are met, rather than always performing the full iterative process. This reduces computational complexity in cases where fewer iterations are sufficient to achieve the required accuracy, while still maintaining data estimation accuracy when needed
Solution Approach 2:
The patent uses parameter changes by optimizing the trainable weights of the receiver algorithm during a training phase using techniques like gradient descent. By pre-training the algorithm with optimal parameters, the actual detection phase requires fewer iterations and less computational effort to achieve high accuracy, effectively reducing the computational complexity during operation
3Reliability
If trainable weights are used in the receiver algorithm, then bit error rate performance improves, but training time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by performing the computationally intensive training of trainable weights in advance, during a separate training phase that occurs before actual data transmission and detection. Once trained, the optimized weights are stored and reused for multiple detection operations, so the time cost of training is amortized over many uses, making the actual detection process fast and efficient
Data Source
AI summary
An apparatus, computer program and method is described including receiving data at a receiver of a communication system, generating an estimate of the data as transmitted by a transmitter of the transmission system (wherein generating the estimate includes a receiver algorithm having at least some trainable weights), generating a refined estimate of the transmitted data, based on said estimate and an error correction algorithm (wherein, in an operational mode, said estimate of the data as transmitted is generated based on the received data and said refined estimate); and generating, in the operational mode, a revised estimate of the transmitted data on each of a plurality of iterations of said generating an estimate of the transmitted data until a first condition is reached.