Extraction method of time-varying signal components based on spi sparsity constraint

Abstract

The invention discloses a time-varying signal component extraction method based on SPI sparse constraints, including: a) SPI sparse performance estimation index and a time-varying signal component estimation paradigm including the SPI index, and b) a time-varying signal component estimation paradigm based on the above paradigm The signal component extraction method includes five independent steps: data preprocessing, system initialization, sparse modeling and waveform dictionary learning, single-time specific component estimation of time-varying signals, and discovery of new heterogeneous target component waveforms. The advantage of the present invention is that, compared with traditional methods, it has the advantage of not relying on statistical stationarity or independence assumptions, can obtain more reliable time-varying signal-specific component dictionary and blind estimation of time-varying signal-specific components, and has unsupervised The ability to discover unknown target components has broad potential applications in the analysis of time-varying signals with complex structures and many unknown or undiscovered components. It is applicable to the improvement of various equipment and algorithms for time-varying signal component analysis and extraction.

Description

technical field [0001] The invention relates to the technical field of time-varying signal component extraction, in particular to a method for constructing a time-varying signal component dictionary and extracting specific components using a brand-new sparse performance index SPI (Sparse Performance Index). Background technique [0002] Blind source separation and blind component extraction and analysis of time-varying signals, especially component extraction without reference and without clear known objects, has always been a difficult problem in the field of digital signal processing and analysis. Due to the lack of accurate estimation of the real components of unknown signals, the processing, analysis and identification of signal data will be limited to varying degrees. Traditional blind source separation or component analysis algorithms generally have pre-assumptions about the real components of the sample set (these assumptions are usually not completely in line with th...

AI Technical Summary

Problems solved by technology

Due to the lack of accurate estimation of the real components of unknown signals, the processing, analysis and identification of signal data will be limited to varying degrees.

Traditional blind source separation or component analysis algorithms generally have pre-assumptions about the real components of the sample set (these assumptions are usually not completely in line with the real situation), such as the independence assumptions of PCA and ICA; in addition, these methods will be affected by the observation data channel. It is difficult to deal with underdetermined and severely underdetermined signal blind estimation problems, and there is an unavoidable problem of aliasing of the components obtained from the analysis

The traditional sparse decomposition algorithm is data-driven and has no pre-assumptions about the true composition of the sample set. However, due to the insufficient precision of the sparse constraint items in engineering implementation, satisfactory results cannot be obtained.

Images