Experimental results: detection and tracking of low SNR sinusoids using real-time LMS and RLS lattice adaptive line enhancers

The authors present a comparison of the detection and tracking performance of adaptive line enhancers (ALEs) implemented with the traditional LMS (least mean square) transversal and the RLS (recursive least squares) lattice algorithms. This comparison is based on experimental results obtained from a real-time custom hardware system using 32-bit IEEE floating point format operating on stationary and nonstationary sinusoids with added broadband noise. Narrowband gain for stationary inputs was measured for the LMS transversal filter and for two forms of the RLS lattice filter. They were found to be the same within reasonable measurement criteria and agreed with the theoretical gain of the LMS ALE. For stationary signals the performance of the RLS lattice exhibits similar variation with filter length and memory fade factor, as does the LMS algorithm. For nonstationary signals, the LMS and the RLS lattice algorithms perform the same for short filters and small FM sweep rates. For longer filters and for higher sweep rates, the RLS lattice exhibits approximately a 3-dB advantage over the LMS algorithm.<
>