|
|
MIT Electrical Engineering and Computer Science
EECS Event |
Monday, April 2, 2001
2:00 PM (refreshments 1:45)
Grier Room, Room 34-401A
EECS Special Seminar
Abstract
Estimation of time-varying parameters is necessary for many applications in communication and signal processing. Conventional adaptive techniques must select a rate of adaptation and, thereby, must tradeoff the error due to misadjustment against the error due to lag in tracking time-varying parameters. Signal-dependent bases provide a means to overcome this tradeoff.
In this presentation, a technique is introduced to estimate parameters in a basis that adapts to signal time variations. We first consider the estimation of linear time-varying (LTV) channels. A basis of smooth local complex exponentials, which are approximate eigenfunctions of underspread LTV channels, is chosen to select variable duration time intervals during which the time variation of the channel is small. Within each interval, estimates of the local transfer function and impulse response are obtained. This estimation technique is able to track LTV channels better than a method that uses time intervals of a fixed duration.
In the second part of the presentation, we use adaptive bases to estimate and track the speed and average received power of a subscriber in a mobile radio system. The use of adaptive bases is motivated by the local stationarity of the received signal. Adaptive bases of smooth local complex exponentials are used to estimate the time-varying Doppler power spectrum and, thereby, identify the Doppler frequencies of the dominant multipath components. Angles of arrival of the dominant multipath components are estimated using antenna arrays. From the time-varying Doppler power spectrum and the angles of arrival, we obtain an estimate of the maximum Doppler frequency, which is proportional to the subscriber speed. This technique is superior to an adaptive averaging method for variable mobile speeds.