UEGK De-Binarization Using Lookup Tables for Faster Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for decoding codewords using the concatenated unary/k-th order Exp-Golomb code (UEGK) scheme require significant hardware, particularly two adders, which can negatively impact the timing path in hardware design.
Innovation Solution
The proposed method involves identifying and processing a codeword's first, second, and third portions to generate an offset using look-up tables or logical operations, eliminating the need for adders by using a look-up table to map the number of successive bits in the second portion to an offset, and adding a predefined value to derive the decoded value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional decoding method using two adders is used, then decoding accuracy is maintained, but hardware complexity increases and timing path deteriorates
Solution Approach 1:
The patent replaces the mechanical addition operation (using adders) with a lookup table-based retrieval operation. Instead of computing the sum of index values through addition, the system pre-computes and stores the mapping between bin string patterns and their corresponding decoded values in a lookup table, then retrieves the result directly, substituting the arithmetic mechanism with a memory access mechanism.
Solution Approach 2:
The patent performs preliminary computation by pre-calculating all possible decoded values and storing them in a lookup table before runtime. The complex addition operations are performed in advance during table generation, and during actual decoding, only simple table lookup and index selection are needed, eliminating the need for adders during the critical decoding path.
2Speed
If conventional decoding method with two adders is used, then complete decoding functionality is achieved, but timing path is negatively impacted
Solution Approach 1:
The patent substitutes the arithmetic addition mechanism with a memory retrieval mechanism. The lookup table stores pre-computed results, allowing the system to replace time-consuming addition operations with faster memory access operations, thereby improving the timing path while reducing the need for arithmetic logic units.
Solution Approach 2:
The patent creates a copy of the decoded value mapping in the form of a lookup table, which can be directly accessed without performing the original computation. This copying approach allows the system to trade off memory usage for computational speed, eliminating adders from the critical timing path.
3Measurement precision
If unary binarization is used, then each codeword is uniquely distinguishable, but codeword length increases significantly
Solution Approach 1:
The patent segments the bin string into multiple portions (first portion, second portion, third portion) and processes each segment to extract index values. This segmentation allows the system to handle large codeword values efficiently by breaking them into manageable components that can be independently decoded and combined through the lookup table mechanism, avoiding the need for long unary codewords.
Solution Approach 2:
The patent transitions from a one-dimensional unary representation to a multi-dimensional structure by dividing the bin string into multiple portions and using a lookup table that maps combinations of these portions to decoded values. This dimensional change allows more compact representation of codewords while maintaining unique distinguishability.
4Length of moving object
If Exp-Golomb binarization is used, then number of bins is reduced, but quick distinction between smaller and larger codewords is lost
Solution Approach 1:
The patent performs preliminary organization of codeword patterns in the lookup table, pre-computing the relationship between bin string patterns and their decoded values. This allows the system to quickly distinguish between smaller and larger codewords during decoding by directly retrieving pre-organized information, compensating for the loss of inherent ordering in Exp-Golomb binarization.
Data Source
AI summary
A method for generating a decoded value from a codeword which is binarized utilizing a concatenated unary/k-th order Exp-Golomb code includes: identifying a first portion of the codeword, a second portion of the codeword and a third portion of the codeword; generating an offset according to the second portion; decoding the third portion to generate an index value; and generating the decoded value by adding the offset and the index value.