Robust signal recovery using the prolate spherical wave functions and maximum correntropy criterion
Autor: | Kit Ian Kou, Cuiming Zou |
---|---|
Rok vydání: | 2018 |
Předmět: |
Signal processing
Mean squared error Noise (signal processing) Mechanical Engineering Speech recognition Aerospace Engineering 020206 networking & telecommunications 02 engineering and technology Prolate spheroid Computer Science Applications Distribution (mathematics) Control and Systems Engineering Special functions Signal Processing Outlier 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Spherical wave functions Algorithm Civil and Structural Engineering Mathematics |
Zdroj: | Mechanical Systems and Signal Processing. 104:279-289 |
ISSN: | 0888-3270 |
DOI: | 10.1016/j.ymssp.2017.10.025 |
Popis: | Signal recovery is one of the most important problem in signal processing. This paper proposes a novel signal recovery method based on prolate spherical wave functions (PSWFs). PSWFs are a kind of special functions, which have been proved having good performance in signal recovery. However, the existing PSWFs based recovery methods used the mean square error (MSE) criterion, which depends on the Gaussianity assumption of the noise distributions. For the non-Gaussian noises, such as impulsive noise or outliers, the MSE criterion is sensitive, which may lead to large reconstruction error. Unlike the existing PSWFs based recovery methods, our proposed PSWFs based recovery method employs the maximum correntropy criterion (MCC), which is independent of the noise distribution. The proposed method can reduce the impact of the large and non-Gaussian noises. The experimental results on synthetic signals with various types of noises show that the proposed MCC based signal recovery method has better robust property against various noises compared to other existing methods. |
Databáze: | OpenAIRE |
Externí odkaz: |