TY - JOUR

T1 - Sequential Amplitude Estimation in Multiuser Communications

AU - Steinberg, Yossef

AU - Poor, H. Vincent

N1 - Funding Information:
Manuscript received April 10, 1992; revised March 17, 1993. This work was supported in part by a Wolfson Postdoctoral Fellowship, and in part by the National Science Foundation under Grant NCR-90-02767. An early version of this paper was presented at the 31st IEEE Conference on Decision and Control, Tucson, AZ, December 1992.

PY - 1994/1

Y1 - 1994/1

N2 - This paper considers the problem of multiuser amplitude estimation, i.e., the problem of estimating the amplitudes of several digital communications signals superimposed in the same channel. This problem is of importance in communications environments such as spread-spectrum radio networks, in which nonorthogonal multiplexing is used. Multiuser amplitude estimation is a critical prerequisite to the optimum demodulation of such signals using, for example, Verdu’s algorithm. Here, a sequential detection-estimation approach is applied to this problem, and several estimation paradigms, including the method of moments and likelihood-based estimators, are considered. The consistency, asymptotic variance, and complexity of these estimators are examined. A new method of constructing a recursive consistent and asymptotically efficient estimation algorithm out of a consistent estimator sequence is also suggested and is applied to the current setup. It is seen that detector-estimators that use these estimators in Verdu’s algorithm result, asymptotically, in (known-amplitude) optimum error probabilities with little relative increase in complexity per demodulated bit.

AB - This paper considers the problem of multiuser amplitude estimation, i.e., the problem of estimating the amplitudes of several digital communications signals superimposed in the same channel. This problem is of importance in communications environments such as spread-spectrum radio networks, in which nonorthogonal multiplexing is used. Multiuser amplitude estimation is a critical prerequisite to the optimum demodulation of such signals using, for example, Verdu’s algorithm. Here, a sequential detection-estimation approach is applied to this problem, and several estimation paradigms, including the method of moments and likelihood-based estimators, are considered. The consistency, asymptotic variance, and complexity of these estimators are examined. A new method of constructing a recursive consistent and asymptotically efficient estimation algorithm out of a consistent estimator sequence is also suggested and is applied to the current setup. It is seen that detector-estimators that use these estimators in Verdu’s algorithm result, asymptotically, in (known-amplitude) optimum error probabilities with little relative increase in complexity per demodulated bit.

KW - Amplitude estimation

KW - multiuser communication

KW - stochastic approximations

UR - http://www.scopus.com/inward/record.url?scp=0028204746&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0028204746&partnerID=8YFLogxK

U2 - 10.1109/18.272451

DO - 10.1109/18.272451

M3 - Article

AN - SCOPUS:0028204746

VL - 40

SP - 11

EP - 20

JO - IEEE Transactions on Information Theory

JF - IEEE Transactions on Information Theory

SN - 0018-9448

IS - 1

ER -