An echo tail canceller based on adaptive interpolated FIR filtering

Abstract

A computationally efficient two-stage echo canceller is proposed. The first stage is a direct form adaptive FIR filter to cancel the main part of the echo response. The second stage cancels the "tail" part of the echo response. The proposed tail canceller is an adaptive version of the interpolated FIR (IFIR) filter in [11-[31. It is shown that the two-stage canceller meets the 60-dB echo cancellation accuracy for various typical loops. Extensive simulations were carried out to study the effects of the main canceller length as well as the interpolator parameters on the performance of the proposed two-stage echo canceller in the U-interface transceiver for basic rate ISDN. It is shown that the proposed structure compares favorably with a recently published two-stage canceller.

High-Performance Low-Cost DFE using IFIR Filters

“…IFIR filters are always stable, can produce sharp frequency responses, and are more robust to coefficient quantization than the conventional FIR filters [6,7]. IFIR filters have been used in many signal processing applications ranging from echo cancellation [8,9] to reduced rank MMSE parameter estimation techniques [10] and have also played a key role in a number of communication systems such as adaptive interference suppression receivers in DS-CDMA systems [11], and equalizer design [9]. This latter work considers however a standard linear equalizer structure that processes a downsampled version of the received data.…”

Analytical model for the mean weight behavior of adaptive interpolated-FIR filters using the constrained filtered LMS algorithm

“…The input signal is a colored noise obtained from a second-order AR process, given by x(n) = a,x(n -1) + u,x(n -2) + u(n) , where u(n) is a white noise process with variance equal to 0.0322; U, = -1.5955, u2 =0.95 ; and the eigenvalue spread is 10. For this example pn,Ly = 0.02 has been experimentally determined.The interpolating factor is L = 2 .Figures 6 and 7follow the same pattern as in Figs 4. and 5.…”

“…To include the constraints in the cost function, the method of Lagrange multipliers is used. Thus as in [9], the constrained cost function is stated as follows minimize E[ e' (ti)] (4) subject to CTW = f ,…”

“…Additionally, AIFIR filters retain most of the desirable features of the traditional AFIR filters. Most of the papers on AIFIR structures found in the open literature concentrate basically on the performance comparison to the ordinary AFIR filters, which have been very important for assessing the potential of these structures [4,5]. However, none of them has presented an accurate model for predicting the mean weight behavior under different implementation conditions.…”

“…This represents a major drawback, mainly due to the large number of arithmetic operations involved for both filtering and weight updating. In those cases, adaptive IFIR (AIFIR) filters may represent an interesting alternative to AFIR filters [4,5]. Additionally, AIFIR filters retain most of the desirable features of the traditional AFIR filters.…”