A low complexity adaptive interpolated FIR echo canceller

Abstract

A novel composite FIR-based filter structure is proposed to reduce the complexity of echo cancellers. The proposed filter consists of an FIR filter, an interpolated FIR (IFIR) filter, and a tap-weight overlapping method. The FIR filter is used to model the short rapidly changing head echo portion while IFIR filter is for the long slowly decaying tail echo. About 55% complexity reduction is obtained compared to a conventional transversal filter while this design inherits stable convergence advantages. Simulation results show that an average 75.1 dB ERLE performance is achieved for the 2320-kbps SDSL system.