Time-Frequency Feature Extraction for Wireless Sensor Data Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wireless Sensor Networks face challenges in reducing data size for efficient transmission, as continuous data transmission consumes significant bandwidth and energy, especially with non-stationary data that varies in statistical and spectral properties, leading to transmission of trivial information.
Innovation Solution
The method involves calculating Wigner Ville Distributions and Renyi entropies for raw data, identifying windows based on entropy thresholds, computing Wigner Ville Spectrum, and classifying Eigen values to reduce data size by clustering relevant categories of events, thereby storing only essential spectral features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If continuous data transmission is performed in Wireless Sensor Networks, then complete information is transmitted, but bandwidth consumption and energy usage increase significantly
Solution Approach 1:
The patent extracts and transmits only the most significant features from sensor data using Wigner Ville Spectrum analysis and Renyi entropy calculation. By identifying and transmitting only relevant spectral features rather than complete raw data, the system maintains information quality while dramatically reducing bandwidth consumption and energy usage in wireless sensor networks
Solution Approach 2:
The patent transforms raw sensor data into a different parameter space using time-frequency analysis (Wigner Ville Distribution) and entropy-based feature selection. This parameter transformation allows the system to represent complex sensor signals compactly using only the most informative features, reducing transmission requirements while preserving essential information
2Quantity of substance
If data compression is applied to non-stationary sensor data, then transmission volume is reduced, but reconstruction distortion increases
Solution Approach 1:
The patent applies Wigner Ville Time-Frequency analysis to transform sensor data from the time domain into a time-frequency domain representation. This dimensional transformation allows non-stationary signals with time-varying spectral properties to be properly analyzed and compressed, capturing both temporal and spectral characteristics without the distortion that plagues traditional compression methods
Solution Approach 2:
The patent performs preliminary feature extraction and selection before compression by calculating Renyi entropies and identifying significant spectral features in advance. This preliminary action ensures that only the most informative features are retained and transmitted, preventing information loss while achieving effective compression of non-stationary sensor data
3Loss of information
If all sensor data is transmitted, then no trivial information is lost, but transmission speed decreases due to large data volume
Solution Approach 1:
The patent extracts only the most significant spectral features from sensor data using Wigner Ville Spectrum analysis and Renyi entropy-based feature selection. By transmitting only these essential features rather than complete raw data, the system achieves fast transmission speeds while maintaining high information quality and eliminating trivial redundant information
Data Source
AI summary
Disclosed is a method and system for reducing data size of raw data. The system may process the raw data for calculating Renyi entropies, Wigner Ville Distributions (WVD's), Wigner Ville Spectrum (WVS) and Renyi divergence. The system may identify a first set of windows followed by a second set of windows while processing the raw data. Further, the system may calculate Eigen values for a Time-Frequency matrix of WVS of the second set of windows. The system may filter the second set of windows based on the Eigen values for preparing a third set of windows. The system prepares clusters of the Eigen values. The system may compute centroids of the clusters of the Eigen values. The system classifies each window of the third set of windows into one of the clusters indicating a relevant category of event identified from the raw data.