Block Floating-Point Compression Using Exponent Token Codes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal compression methods fail to efficiently manage dynamic range and bit representation for signal samples, particularly in applications where signal amplitudes fluctuate over time, leading to suboptimal data storage and transmission efficiency.
Innovation Solution
The method involves grouping signal samples into blocks, determining a common exponent and adjusting the number of bits for each mantissa based on the block exponent, and encoding these values to form compressed data, allowing for flexible output data rate or quality metric control during compression and decompression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fixed-point or floating-point representation is used for signal samples, then dynamic range is limited, but maintaining accuracy across varying amplitudes requires more bits per sample
Solution Approach 1:
The signal is divided into blocks of samples, and each block is processed independently with its own exponent determination. This segmentation allows the system to adapt to local amplitude variations without requiring high precision for all samples uniformly, reducing total bits while maintaining accuracy where needed
Solution Approach 2:
The patent changes the representation parameters by using block-floating-point format where the exponent is shared across a block of samples rather than being fixed for all samples or varying per sample. This parameter change allows efficient bit allocation that adapts to the actual signal dynamics in each block
2Productivity
If variable bit allocation is used for mantissas based on signal amplitude, then compression efficiency improves, but encoding complexity increases
Solution Approach 1:
The patent merges the exponent information into a shared block-level parameter rather than encoding it separately for each sample. This combining of the exponent determination across multiple samples reduces the total number of encoding operations required while maintaining variable precision where needed
Solution Approach 2:
The system dynamically determines the exponent for each block based on the actual signal amplitudes present in that block. This dynamic adaptation allows the encoder to optimize bit allocation in real-time without requiring complex pre-computation or lookup tables
3Duration of action of moving object
If common exponent is determined for each block based on largest magnitude sample, then dynamic range is expanded, but precision is lost for smaller amplitude samples
Solution Approach 1:
The patent applies local quality by allowing each block to have its own locally-determined exponent based on the maximum amplitude in that specific block. This local adaptation ensures that the dynamic range is optimized for each block's characteristics while maintaining adequate precision for all samples within that block
4Measurement precision
If more bits are allocated to represent signal samples with varying amplitudes, then accuracy is maintained, but data storage and transmission requirements increase
Solution Approach 1:
The patent changes the fundamental representation parameters from fixed-point or per-sample floating-point to block-floating-point format. This parameter change enables the system to maintain signal accuracy while reducing data volume by sharing exponent information across multiple samples within each block
Data Source
AI summary
A method and apparatus for compressing signal samples uses block floating point representations where the number of bits per mantissa is determined by the maximum magnitude sample in the group. The compressor defines groups of signal samples having a fixed number of samples per group. The maximum magnitude sample in the group determines an exponent value corresponding to the number of bits for representing the maximum sample value. The exponent values are encoded to form exponent tokens. Exponent differences between consecutive exponent values may be encoded individually or jointly. The samples in the group are mapped to corresponding mantissas, each mantissa having a number of bits based on the exponent value. Removing LSBs depending on the exponent value produces mantissas having fewer bits. Feedback control monitors the compressed bit rate and/or a quality metric. This abstract does not limit the scope of the invention as described in the claims.