Neural Network Soft Detection for Read Channel Error Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data detection methods in read channels of data transmission and storage devices, such as PRML and LDPC, face limitations in error rate performance and complexity, necessitating an improvement in signal processing techniques.
Innovation Solution
The implementation of a neural network-based soft information detector, specifically an artificial neural network (ANN) detector, which determines symbol probabilities for a SOVA detector in an iterative detection loop, enhancing error rate performance and reducing noise correlation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional PRML and LDPC detection schemes are used, then the read channel can maintain relatively simple processing, but error rate performance is limited
Solution Approach 1:
The patent replaces traditional mechanical/signal-processing-based detection methods (PRML, LDPC) with a neural network-based detection system. The neural network detector uses machine learning models trained on channel characteristics to perform detection, substitution of conventional signal processing algorithms with AI-based processing, achieving improved error rate performance while managing complexity through the adaptive nature of neural networks
Solution Approach 2:
The patent changes the operational parameters of the detection system by using soft information (probabilities) instead of hard decisions. The neural network outputs probability distributions over possible symbol values, which are then used by the iterative decoder. This parameter change from binary decisions to probabilistic information enables better error rate performance
2Reliability
If iterative detection with soft information is implemented, then error rate performance improves, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network detector offline using training sequences and channel characteristics. The neural network weights are pre-optimized to match the specific channel conditions, so that during actual operation, the detector can make accurate predictions without requiring complex real-time computations. This preliminary training phase transfers the computational burden from runtime to setup time
Solution Approach 2:
The patent implements feedback through the iterative detection loop where the neural network detector outputs soft information to the decoder, which then feeds back extrinsic information to refine the detection. This feedback mechanism allows the system to progressively improve detection accuracy over multiple iterations, achieving better error rate performance while distributing computational load
Data Source
AI summary
Example systems, read channels, and methods provide bit value detection from an encoded data signal using a neural network soft information detector. The neural network detector determines a set of probabilities for possible states of a data symbol from the encoded data signal. A soft output detector uses the set of probabilities for possible states of the data symbol to determine a set of bit probabilities that are iteratively exchanged as extrinsic information with an iterative decoder for making decoding decisions. The iterative decoder outputs decoded bit values for a data unit that includes the data symbol.