LL Logo




Estimation of a Jammed Multipath Channel Delay Profile Using Superresolution Methods and a Random Path Decorrelation Technique

Frantz Bouchereau, David Brady, and Colin Lanzl
Northeastern University
409 Dana/Northeastern University
Boston, MA 02115
tel: (617) 373-5400
email: brady@cdsp.neu.edu

Abstract High resolution echo location in multipath channels is a seminal issue in wireless indoor source localization, and in underwater acoustic applications. In this work we extend a superresolution pseudo-noise sequence correlation method (SPM) to the time delay estimation of a jammed multipath channel. This method is based on running the MUSIC (Multiple Signal Classification) algorithm on the covariance matrix of the low resolution delay profile vector. Since MUSIC assumes a decorrelated signal covariance matrix, successful decorrelation of the inherently coherent multipath echoes becomes a critical issue. We show that previously proposed path decorrelation techniques will fail in certain jammed multipath environments and will be, in many cases, computationally expensive. We propose an improved variation of this decorrelation technique, called Random Frequency Smoothing, which will need fewer computations. We investigate the effects of a narrowband jammer in the SPM algorithm and propose a practical, simple way to null its effect. We also propose a transmission method that shortens the transmission period and reduces channel variations during signal acquisition. Finally, we show the results of several underwater acoustic transmission experiments to which we applied the proposed SPM algorithm. These experiments were performed in a water pond and in deep water ocean. A 4 kHz chip rate PN-sequence signal was transmitted on channels of approximately 10 kHz BW and 23 dB SNR. We consider the following practical issues: phase rotations, limited number of time snapshots, unknown number of paths, modest signal-to-noise ratio, and nonlinearities of power amplifiers. SPM performance was compared to correlation peak detection using barker sequences of equal and twice the PN-sequence bandwidth. SPM outperformed the correlation delay profiles even for the double bandwidth case. Paths separated by as close as 0.08 ms were resolved.

Presentation (pdf format)



LL Logo Disclaimer

Direct comments and questions to: webmaster@ll.mit.edu

MIT Lincoln Laboratory. All rights reserved.