Gradient estimation for sensitivity analysis and adaptive multiuser interference rejection in code-division multiple-access systems

In this paper, we consider a direct-sequence code-division multiple-access (DS-CDMA) system in the framework of a discrete-event dynamic system (DEDS) in order to optimize the system performance. Based on this formulation, we develop infinitesimal perturbation analysis (IPA) for estimating the sensi...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transactions on communications Ročník 45; číslo 7; s. 848 - 858
Hlavní autoři: Mandayam, N.B., Aazhang, B.
Médium: Journal Article
Jazyk:angličtina
Vydáno: IEEE 01.07.1997
Témata:
Detectors
Error correction
Frequency estimation
Interference
Multiaccess communication
Power control
Sensitivity analysis
Stochastic processes
Stochastic systems
System performance
ISSN:0090-6778
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this paper, we consider a direct-sequence code-division multiple-access (DS-CDMA) system in the framework of a discrete-event dynamic system (DEDS) in order to optimize the system performance. Based on this formulation, we develop infinitesimal perturbation analysis (IPA) for estimating the sensitivity of the average probability of bit error to factors ranging from near-far effects to imperfections in power control. The above estimates are shown to be unbiased, and this technique is then further incorporated into a stochastic gradient algorithm for achieving adaptive multiuser interference rejection for such systems, which is also subject to frequency nonselective slow fading. We use an IPA-based stochastic training algorithm for developing an adaptive linear detector with the average probability of error being the minimization criterion. We also develop a practical implementation of such an adaptive detector where we use a joint estimation-detection algorithm for minimizing the average probability of bit error. A sequential implementation that does not require a stochastic training sequence or a preamble is also developed.
Bibliografie:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Feature-1
content type line 23
ISSN:0090-6778
DOI:10.1109/26.602590