Multimodal Signal Compression Using Wavelets and Autoencoders
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Transmitting large amounts of high-resolution raw data from power systems, such as those in the PowerGenome architecture, over networks is inefficient due to bandwidth limitations, making it unfeasible with current communication technologies.
Innovation Solution
The method and system utilize data featurization and compression techniques, specifically combining sparse wavelet transforms and autoencoders, to reduce the data size by 3 or more orders of magnitude. This approach records and transmits only the differences between subsequent signals, reducing the need for extensive computing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution raw sensor data is transmitted over the network, then measurement precision is improved, but data transmission bandwidth requirements increase significantly
Solution Approach 1:
The patent extracts and transmits only the essential features and differences from the raw sensor data rather than the complete high-resolution signals. By recording only the difference between subsequent signals (Δx) and applying featurization techniques, the system separates the critical information from the redundant data, achieving both precision and compression.
Solution Approach 2:
The patent transforms the data from its original high-resolution form into a compressed representation by changing parameters such as recording differences between consecutive measurements rather than absolute values. This parameter transformation reduces data volume while preserving the essential information needed for system optimization.
2Reliability
If large amounts of raw data are collected and transmitted, then system optimization accuracy is improved, but network bandwidth consumption increases
Solution Approach 1:
The patent extracts only the necessary information for system optimization by recording differences between subsequent signals and applying featurization. This extraction process removes redundant data while retaining the essential patterns and relationships needed for accurate system optimization, thereby reducing transmission volume without compromising optimization reliability.
Solution Approach 2:
The patent performs preliminary processing of the sensor data at the source by calculating differences between consecutive measurements and applying compression techniques before transmission. This preliminary action reduces the data volume that needs to be transmitted over the network while preserving the information necessary for accurate system optimization.
3Adaptability or versatility
If comprehensive signal information is collated from multiple devices, then system analysis capability is improved, but computing power requirements increase
Solution Approach 1:
The patent extracts essential features from comprehensive multi-device signal information through featurization and difference calculation, retaining the critical patterns needed for system analysis while removing redundant details. This extraction enables comprehensive system analysis with reduced computing power requirements at edge devices.
Solution Approach 2:
The patent transforms comprehensive signal data into a compressed feature space by changing parameters such as recording differences between measurements and applying wavelet transforms. This parameter transformation maintains the analytical capabilities needed for system optimization while significantly reducing the computing power required for data processing.
4Productivity
If high transmission bandwidth is provided to handle raw data, then data transfer efficiency is improved, but system cost and complexity increase
Solution Approach 1:
The patent extracts and transmits only the essential difference information and features from raw sensor data, reducing the data volume that requires high-bandwidth transmission. This extraction approach achieves efficient data transfer using standard network bandwidths without requiring expensive high-capacity communication infrastructure.
Solution Approach 2:
The patent changes the parameters of the transmitted data from high-resolution raw values to compressed difference representations and feature vectors. This parameter transformation enables efficient data transfer over standard network bandwidths, avoiding the need for expensive high-capacity communication systems while maintaining data utility for system optimization.
Data Source
AI summary
A data transmission and compression method arranged to compress and transmit data between an edge device and a remote server. The method includes collecting data at the edge device, wherein the data is attributed to a plurality of signal signatures, generating a data matrix at the edge device, and transforming the data matrix, wherein transforming the data includes using a wavelet transform. The method further includes compressing the data, wherein compressing the data includes utilising an autoencoder, and transmitting the encoded compressed data to the remote server via a communication channel. The method further includes, at the remote server, decompressing the data utilising an autoencoder, reconstructing the signal signatures using an inverse wavelet transform, and storing the reconstructed data signatures in a datastore on the remote server.