System and method for signal pre-processing based on data driven models and data dependent model transformation

A system for signal pre-processing based on a plurality of data driven models and a data dependent model transformation, comprising:

• a memory storing instructions;
• one or more communication interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
• receive a raw signal as an input;
• learn a set of representational basis from the received raw signal, wherein the set of representational basis comprises a plurality of orthonormal vectors;
• compute a domain dependent time-frequency energy distribution pattern by averaging of a plurality of dictionary atoms, wherein the plurality of dictionary atoms is a linear signal space;
• select at least one orthonormal vector from the plurality of orthonormal vectors based on the computed domain dependent time-frequency energy distribution pattern, wherein the selection of the at least one orthonormal vector corresponds to the plurality of dictionary atoms;
• determine a structure of the plurality of dictionary atoms, wherein the structure of the plurality of dictionary atoms corresponds to a graph structure represented as a Laplacian matrix (L);
• integrate the graph structure as a structure of the set of representational basis to obtain a reconfigured data model;
• reconstruct, using the reconfigured data model, the raw signal, to obtain a denoised signal;
• determine a structure of the reconstructed raw signal based on graph structured dictionary atoms, wherein
• constraints on an optimization problem to a specific spectral and a topological structure of the reconstructed raw signal;
• perform a mapping between a process, the structure, a property, and a performance of the denoised signal, wherein
• the process of the denoised signal is a portfolio of algorithms,
• the portfolio of algorithms includes learning the plurality of dictionary atoms, and re-structuring the denoised signal based on spectral and topological properties,
• the structure of the denoised signal defines the property of the denoised signal in terms of accuracy, consistency, completeness, validity, and density,
• the performance is related to estimation, detection, clustering and classification procedures of the denoised signal, and
• the property of the denoised signal is related to inferences on the denoised signal;
• determine, based on the mapping, whether an output performance of the signal pre-processing of the raw signal is matching with a threshold performance;
• execute, based on a mismatch between the output performance of the signal pre-processing and the threshold performance, a process-structure-property-performance map;
• and
• determine, based on the execution of the process-structure-property-performance map, a modification required in a process parameters of the signal pre-processing to match the output performance of the signal pre-processing and the threshold performance.

One or more non-transitory machine-readable information storage mediums comprising one or more instructions, which when executed by one or more hardware processors cause:

• receiving a raw signal as an input;
• learning a set of representational basis from the received raw signal, wherein the set of representational basis comprises a plurality of orthonormal vectors;
• computing a domain dependent time-frequency energy distribution pattern by averaging of a plurality of dictionary atoms, wherein the plurality of dictionary atoms is a linear signal space;
• selecting at least one orthonormal vector from the plurality of orthonormal vectors based on the computed domain dependent time-frequency energy distribution pattern, wherein the selection of the at least one orthonormal vector corresponds to the plurality of dictionary atoms;
• determining a structure of the plurality of dictionary atoms, wherein the structure of the plurality of dictionary atoms corresponds to a graph structure represented as a Laplacian matrix (L);
• integrating the graph structure as a structure of the set of representational basis to obtain a reconfigured data model;
• reconstructing, using the reconfigured data model, the raw signal to obtain a denoised signal;
• determining a structure of the reconstructed raw signal is determined based on graph structured dictionary atoms, wherein
• constraints on an optimization problem corresponds to a specific spectral and a topological structure of the reconstructed raw signal;
• performing a mapping between a process, the structure, a property, and a performance of the denoised signal, wherein
• the process of the denoised signal is a portfolio of algorithms,
• the portfolio of algorithms includes learning the plurality of dictionary atoms, and re-structuring the denoised signal based on spectral and topological properties,
• the structure of the denoised signal defines the property of the denoised signal in terms of accuracy, consistency, completeness, validity, and density,
• the performance is related to estimation, detection, clustering and classification procedures of the denoised signal, and
• the property of the denoised signal is related to inferences on the denoised signal;
• determining, based on the mapping, whether an output performance of the signal pre-processing of the raw signal is matching with a threshold performance;
• executing, based on a mismatch between the output performance of the signal pre-processing and the threshold performance, a process-structure-property-performance map;
• and
• determining, based on the execution of the process-structure-property-performance map, a modification required in a process parameters of the signal pre-processing to match the output performance of the signal pre-processing and the threshold performance.