A steady-state Kalman predictor-based filtering strategy for non-overlapping sub-band spectral estimation

Area of Science:

• Signal Processing
• Spectral Analysis
• Digital Filtering

Background:

• Spectral overlap in sub-band spectral estimation complicates analysis.
• Finite-length sequences weaken high-order filter performance due to zero-padding in convolutions.
• Existing algorithms like nonlinear least squares spectral analysis face high computational costs.

Purpose of the Study:

• To suppress spectral overlap in sub-band spectral estimation.
• To reduce the computational complexity of spectral estimation algorithms.
• To improve the effectiveness of spectral analysis methods.

Main Methods:

• Proposed an extrapolation-based filtering strategy to substitute zeros in convolution operations.
• Employed a steady-state Kalman predictor for linearly-optimal extrapolation.
• Applied typical spectral analysis methods to validate the strategy.

Main Results:

• Demonstrated the recovery of high-order filter's spectral overlap suppression ability.
• Showcased a significant decrease in computational complexity for spectral estimation.
• Validated the effectiveness of the proposed strategy through comparative analysis.

Conclusions:

• The extrapolation-based filtering strategy effectively suppresses spectral overlap.
• The proposed method enhances spectral estimation accuracy and reduces computational load.
• This approach offers a valuable improvement for spectral analysis applications.