Abstract
Adaptive filters, with their efficiency and simplicity, have been used successfully in various ultrasonic NDT signal processing contexts. Of these, the adaptive deconvolution with the conven- tional least-mean-squares (LMS) adaptive filter has improved time resolution. However; the conver- gence speed of LMS is restricted by the eigenvalue spread of the input correlation matrix. This paper explores the potential of other adaptive algorithms, namely, normalized least-mean-squares (NLMS), recursive least squares (RLS) and QR-decomposition-based RLS (QR-RLS) to handle the deconvo- lution of ultrasonic NDT signals and compare their performances with that of the conventional LMS algorithm. Furthermore, the mean square error (MSE) behavior in the different adaptive filtering algorithms for ultrasonic NDT signals deconvolution is briefly introduced. Experiments results are explained by graphs and discussed based on the performance criteria. The proposed methods enhanced the resolution quality, offering more alternatives for this application according to specific case requirements.
Original language | English |
---|---|
Pages (from-to) | 285-290 |
Journal | Russian Journal of Nondestructive Testing |
Volume | 48 |
Issue number | 5 |
DOIs | |
Publication status | Published - May 2012 |
MoE publication type | A1 Journal article-refereed |
Funding
This study was supported by the Ministry of Knowledge Economy (MKE) through the Regional Inno vation Centre program.
Keywords
- Adaptive deconvolution
- Adaptive LMS
- Normalized LMS
- QR- decomposition-based RLS
- Recursive least-squares