Machine-Learning Data Reading for Interference-Resilient Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage and retrieval systems face challenges with interference in transduction processes, leading to inefficiencies in encoding and decoding data, particularly in high-throughput and high-density optical storage.
Innovation Solution
The use of machine-learning techniques, specifically convolutional neural networks (CNNs), to optimize the encoding and decoding processes in data storage systems, allowing for improved data retrieval and storage efficiency by directly processing component images without intermediate computations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional transduction processes are used for data storage and retrieval, then data can be stored and retrieved, but interference occurs during the process leading to reduced accuracy and efficiency
Solution Approach 1:
The patent replaces traditional optical transduction mechanisms with machine-learning-based computational processing. Instead of relying on physical transduction processes that are susceptible to interference, the system uses CNNs to directly analyze optical storage medium images and retrieve data, eliminating the harmful interference effects inherent in conventional transduction methods.
Solution Approach 2:
The patent optimizes multiple parameters including laser power, exposure time, and CNN processing parameters to maximize signal quality and minimize interference. By dynamically adjusting these parameters based on the specific storage medium and conditions, the system achieves higher retrieval accuracy while reducing the impact of harmful interference factors.
2Productivity
If conventional encoding and decoding methods are used, then data can be processed, but the process is inefficient and prone to errors in high-throughput systems
Solution Approach 1:
The patent replaces conventional algorithmic encoding and decoding with deep learning-based approaches. Convolutional neural networks are trained to perform both encoding and decoding functions, achieving higher throughput and lower error rates compared to traditional methods. The CNN-based system processes data more efficiently by learning optimal transformations directly from training data rather than following rigid algorithmic procedures.
Solution Approach 2:
The machine-learning model learns to correct its own errors and optimize its performance through training on labeled data. The system automatically adjusts its internal parameters and processing strategies to minimize encoding and decoding errors, enabling self-improving performance without external intervention.
3Ease of operation
If intermediate computations are performed in the decoding process, then data can be decoded, but the process becomes more complex and less efficient
Solution Approach 1:
The patent merges multiple separate processing steps (image acquisition, preprocessing, decoding, and error correction) into a single integrated CNN model. This unified approach eliminates the need for separate intermediate computation steps, simplifying the overall process while maintaining or improving decoding performance. The CNN performs all necessary operations in one cohesive computational framework.
Solution Approach 2:
The patent replaces complex multi-step computational procedures with a single neural network inference operation. The CNN model encapsulates all necessary processing logic within its architecture and trained parameters, allowing the system to perform decoding without explicit intermediate computation steps, thereby reducing operational complexity.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Examples are disclosed that relate to reading stored data. The method comprises obtaining a representation of a measurement performed on a data-storage medium, the representation being based on a previously recorded pattern of data encoded in the data-storage medium in a layout that defines a plurality of data locations. The method further comprises inputting the representation into a data decoder comprising a trained machine-learning function, and obtaining from the data decoder, for each data location of the layout, a plurality of probability values, wherein each probability value is associated with a corresponding data value and represents the probability that the corresponding data value matches the actual data value in the previously recorded pattern of data at a same location in the layout.