Repository logo
 

Interpolation for de-Dopplerisation

Accepted version
Peer-reviewed

Type

Article

Change log

Authors

Graham, WR 

Abstract

'De-Dopplerisation' is one aspect of a problem frequently encountered in experimental acoustics: deducing an emitted source signal from received data. It is necessary when source and receiver are in relative motion, and requires interpolation of the measured signal. This introduces error. In acoustics, typical current practice is to employ linear interpolation and reduce error by over-sampling. In other applications, more advanced approaches with better performance have been developed. Associated with this work is a large body of theoretical analysis, much of which is highly specialised. Nonetheless, a simple and compact performance metric is available: the Fourier transform of the `kernel' function underlying the interpolation method. Furthermore, in the acoustics context, it is a more appropriate indicator than other, more abstract, candidates. On this basis, interpolators from three families previously identified as promising --- piecewise-polynomial, windowed-sinc, and B-spline-based --- are compared. The results show that significant improvements over linear interpolation can straightforwardly be obtained. The recommended approach is B-spline-based interpolation, which performs best irrespective of accuracy specification. Its only drawback is a pre-filtering requirement, which represents an additional implementation cost compared to other methods. If this cost is unacceptable, and aliasing errors (on re-sampling) up to approximately 1% can be tolerated, a family of piecewise-cubic interpolators provides the best alternative.

Description

Keywords

Doppler effect, Acoustic signal processing, Acoustic measurement

Journal Title

Journal of Sound and Vibration

Conference Name

Journal ISSN

0022-460X
1095-8568

Volume Title

422

Publisher

Elsevier BV