Sparse Dictionary Compression for Low-Loss Data Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data compression methods for large data transmission in communication scenarios, such as point cloud and AI model data transmission, result in low compression rates and significant data loss.
Innovation Solution
Decompose data into dictionary and sparse matrices to perform effective data compression, using location indication information and element sequences to express data, thereby improving compression efficiency and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If scalar quantization or vector quantization is used for data compression, then transmission resources are saved and transmission delay is reduced, but compression rate is low and data loss is large
Solution Approach 1:
The patent segments the original data into multiple subdata blocks, and each block is independently compressed using dictionary learning. This segmentation allows for more granular control over the compression process, enabling higher compression rates while maintaining data quality by preserving important features in each segment separately.
Solution Approach 2:
The patent transforms the data representation by changing from traditional scalar/vector quantization parameters to dictionary learning parameters (dictionary matrix and sparse coefficient matrix). This parameter transformation enables more efficient compression by capturing data characteristics in a more compact form, achieving both high compression rate and low data loss.
2Productivity
If traditional quantization methods are used, then compression is achieved, but compression rate remains low
Solution Approach 1:
The patent changes the compression parameters from fixed quantization steps to adaptive dictionary learning parameters. By learning the underlying structure of the data, the dictionary matrix and sparse coefficients achieve much higher compression rates, reducing the quantity of data to be transmitted while preserving essential information.
Solution Approach 2:
The patent performs preliminary dictionary learning and training before the actual compression process. This preliminary action creates an optimized dictionary matrix that captures the essential features of the data, enabling subsequent compression operations to achieve high compression rates without losing critical information.
Data Source
AI summary
The present disclosure relates to data compression and transmission methods, apparatuses, devices, and storage mediums. In one example method, a first communication apparatus decomposes an ith piece of subdata Yi in M pieces of subdata into a dictionary matrix Di and a sparse matrix Xi, where the sparse matrix Xi is used to express the ith piece of subdata Yi based on the dictionary matrix Di, the M pieces of subdata are included in to-be-transmitted first data, M is an integer greater than or equal to 1, and i is an integer greater than or equal to 1 and less than or equal to M. The first communication apparatus performs data compression based on the dictionary matrix Di and the sparse matrix Xi, to obtain compressed data Zi of the ith piece of subdata Yi; and then sends the compressed data Zi to a second communication apparatus.