Universal Multi-Sensor Data Compression via Divisor-Based Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing IoT devices face challenges in efficiently compressing diverse sensor data due to resource constraints, particularly in memory, computation power, and energy, and lack a unified data compression encoder that can handle multi-modality sensor data effectively.
Innovation Solution
A method utilizing divisor-based signal decomposition with sparse matrices and entropy coding to compress signals, reducing redundancy and resource usage, applicable to various types of signals through multi-level decomposition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If lossy data compression methods are used to achieve better compression gain, then compression ratio is improved, but reconstructed signal quality deteriorates
Solution Approach 1:
The patent transforms the signal using a divisor-based transformation matrix that changes the representation parameters of the signal. By selecting appropriate divisors and constructing transformation matrices based on signal length divisors, the method achieves efficient compression while maintaining signal fidelity through parameter optimization rather than simple lossy approximation.
Solution Approach 2:
The patent segments the signal transformation process into multiple levels using a multi-level decomposition structure. The transformation matrix is constructed by combining basis matrices at different levels, allowing the signal to be processed hierarchically and compressed efficiently while preserving important signal characteristics at each level.
2Adaptability or versatility
If multiple miniaturized sensors are integrated into an IoT device to enable novel applications, then sensor data volume increases, but computation and memory resources are consumed
Solution Approach 1:
The patent develops a universal compression framework that can handle multiple types of sensor data (accelerometer, gyroscope, magnetometer, barometer) using the same transformation approach. The divisor-based transformation matrix and multi-level decomposition structure provide a unified solution that adapts to different sensor modalities without requiring separate processing pipelines, thereby reducing overall system complexity.
Solution Approach 2:
The method changes the computational parameters by using divisor-based transformation matrices that are constructed based on the signal length and its divisors. This approach reduces the computational complexity compared to traditional methods by optimizing the transformation process according to the specific characteristics of each sensor signal.
3Productivity
If traditional wavelet transform is used for signal compression, then compression is achieved, but computational complexity increases due to full decomposition
Solution Approach 1:
The patent applies partial decomposition by selecting only certain divisors and levels based on the signal characteristics and compression requirements. Instead of performing complete multi-level decomposition at all possible levels, the method selectively applies transformation at specific levels using the divisor-based approach, thereby reducing computational complexity while maintaining compression effectiveness.
Solution Approach 2:
The transformation process is segmented into multiple independent levels, each handled by separate transformation matrices. This segmentation allows the computation to be distributed and optimized at each level independently, reducing the overall computational burden compared to monolithic transformation approaches.
Data Source
Figure 1a~1b
Figure 1c~2c
Figure 3
AI summary
The present disclosure regards a method for compression of a signal, of length, comprising data, the method comprising the steps of at least for a first level of decomposition, transforming the signal based on a first sparse matrix, of size × formed by an invertible matrix, representing a divisor-based signal decomposition for a first divisor to obtain a first vector of transformation coefficients, wherein =, and wherein the first vector comprises an initial component and detail components,, …,. The method further comprises the step of at least for a second level of decomposition, transforming the initial component based on a second sparse matrix, of size ×, where =, formed by an invertible matrix, representing a divisor-based signal decomposition for a second divisor, to obtain a second vector of transformation coefficients wherein =, and wherein the second vector comprises an initial component and detail components,, …,. After the transformation coefficients of the second vector of transformation coefficients,,, …, and detail components of the first vector,, …,, are quantized to obtain a quantized transformation vector and the signal is compressed based on the quantized transformation vector.