Root cepstral filtering for improved system identification
Root cepstral filtering for improved system identification
- Author(s):
- DOI: 10.1049/cp:20060639
For access to this article, please select a purchase option:
Buy conference paper PDF
Buy Knowledge Pack
IET members benefit from discounts to all IET publications and free access to E&T Magazine. If you are an IET member, log in to your account and the discounts will automatically be applied.
2nd IET International Conference on Intelligent Environments (IE 06) — Recommend this title to your library
Thank you
Your recommendation has been sent to your librarian.
- Author(s): Source: 2nd IET International Conference on Intelligent Environments (IE 06), 2006 page ()
- Conference: 2nd IET International Conference on Intelligent Environments (IE 06)
- DOI: 10.1049/cp:20060639
- ISBN: 0 86341 663 2
- Location: Athens, Greece
- Conference date: 5-6 July 2006
- Format: PDF
Phase information is seen to be important and decisive for the performance of a variety of signal processing tasks, including biometric data processing and robotics/navigational applications which require correct localization of objects or events. Existing approaches to system identification ignore phase information, under the minimum-phase assumption; yet, it is argued that improved performance can be obtained if phase information is retained. Modelling the impulse response of an unknown system from the observed output signal comprised of periodic input excitations, has traditionally been undertaken in the time domain. A novel method for system identification is proposed here, by undertaking the analysis in the frequency domain and utilizing the properties of the root cepstrum, which is a phase-retaining coefficient. We are able to demonstrate, that by warping the z-plane, the system impulse response will be separated from the problematic excitation. Hence, by moving the excitation zeros 'away' from the unit circle, the system impulse response may be extracted and modelled. Potential applications of our method include processing of biometric signals, where improved speaker identification/verification depends critically on an accurate vocal tract model of the speaker, as well as robotic/autonomous navigation applications, where correct automatic localization of items and/or events depends on the signal or system phase information. (5 pages)
Inspec keywords: speaker recognition; biometrics (access control); filtering theory; signal processing; identification; cepstral analysis
Subjects: Signal processing theory; Filtering methods in signal processing; Simulation, modelling and identification
Related content
content/conferences/10.1049/cp_20060639
pub_keyword,iet_inspecKeyword,pub_concept
6
6