Botond Sandor Kirei

Modified NLMF adaptation of Volterra filters used for nonlinear acoustic echo cancellation

When dealing with adaptive nonlinear filters used in acoustic echo cancellation, the tradeoff between convergence rate and steady-state error is an important issue. This tradeoff is favorably addressed here using a modified version of the normalized least-mean-fourth (NLMF) algorithm applied to an adaptive second-order Volterra structure. A convergence rate improvement is obtained at the same steady-state error by amending the step size of the conventional NLMF algorithm using a new step-size function that depends on the norm of an error vector of a certain length. The efficiency of the proposed method is compared with that of the conventional adapted second-order Volterra filter for nonlinear acoustic echo cancellation in terms of echo return loss enhancement (ERLE). From simulations conducted for input signals with different probability density functions, the new approach is shown to outperform the normalized least-mean-square (NLMS) second-order Volterra filter in terms of the convergence rate of the same steady-state error. The proposed technique was also tested in acoustic scenarios with impulse noise and separately with distinct local signal powers. The performance of the method was also compared to that of the traditional NLMF algorithm and of the variable NLMF (XE-NLMF) algorithm. Highlights? A modified NLMF algorithm is proposed in nonlinear AEC using second-order Volterra structures. ? The behavior of the proposed method is supported by the associated cost function. ? The modified NLMF update procedure consists in the normalization of a length-q error vector. ? A dynamic computation of q allows convergence rate improvement for the same steady-state error. ? A comparison with similar adaptation procedures found in the literature is conducted.

Echo cancelling using adaptive algorithms

In the paper, an echo canceller based on a system identification scheme with adaptive algorithms is presented. The algorithms consider FIR filters with the taps chosen to minimize an error signal derived from the system according to a stochastic gradient-based method. The following adaptive filters are discussed and simulated in Matlab: classical LMS and its variants: NLMS, VSLMS, VSNLMS and RLS. The echo was simulated by adding several delayed and attenuated replica to a vocal input. The comparison between the algorithms was made taking into account the mean-square error (MSE) and average attenuation (AV), as well as the required number of multiplications.

Blind Frequency-Selective I/Q Mismatch Compensation Using Subband Processing

This brief proposes a new approach to frequency-selective in-phase (I)/quadrature (Q) mismatch compensation, based on subband processing: The spectrum of the signal received by the digital baseband processor-a combination of wanted and image signals-is split into subbands so that, within each subband, the I/Q mismatch can be considered constant with respect to frequency. Therefore, standard compensation algorithms, devised for frequency-invariant I/Q mismatch, can be applied to compensate the signal within each subband; finally, the outputs of each subband processing path are recombined to yield the wanted signal. This procedure is demonstrated by a case study that describes an I/Q mismatch compensation algorithm particularly well suited for the proposed method and proves that, at least for certain I/Q imbalances, the efficiency of compensation can be significantly improved even if only a small number of subbands are employed. The resulting frequency-selective I/Q mismatch compensation solution is suitable for wideband multistandard receivers.

Comparison of different experimental methods for the assessment of the room’s acoustics

The paper presents the acoustics analysis of three different enclosed spaces. These spaces (rooms) have different geometrical shapes and sizes and serve for different purposes. The early decay time, reverberation time, clarity and center time are evaluated with Dirac, WinMLS, Aurora and Caracad software using simple, low-cost equipment. The listed acoustic parameters were determined using linear sine sweep and impulsive sources. Comparisons between experimental measurements, simulations and analytic results were done. The room impulse response measurement proved to be the most reliable method to evaluate the properties of different rooms even if the measurements are perturbed by non-idealities of the low-cost equipment.

HDL implementation of the variable step size N-LMS adaptive algorithm

This paper proposes a Verilog implementation of the normalized least mean square (NLMS) adaptive algorithm, having a variable step size. The envisaged application is the identification of an unknown system. First the convergence of derived LMS algorithms was analyzed in a Simulink application. The NLMS was chosen for its good convergence and stability. Next a HDL implementation was developed for a 4th order NLMS adaptive filter. ModelSim simulations results altogether with plots obtained in Matlab prove the good behavior of the NLMS algorithm, making it an important competitor for the conventional LMS.

Analysis of Lossless Differential Microstrip Line

In this paper several aspects regarding the analysis of lossless differential microstrip lines are presented. Starting from the physical structure of a coupled microstrip transmission line, the per-unit-length parameters were computed. Then a PSPICE model for a three-conductor transmission line was developed and tested in time and frequency domain. For the exact analysis a system of partial differential equations was set up and a general solution was given.

Adaptive algorithms for double-talk echo cancelling

In the paper, an echo canceller based on a system identification scheme with adaptive algorithms is presented. The algorithms consider FIR filters with the taps chosen to minimize an error signal derived from the system according to a stochastic gradient-based method. The following adaptive filters are discussed and simulated in Matlab: classical LMS and its variants: NLMS, VSLMS, VSNLMS and RLS. The echo was simulated by adding several delayed and attenuated replica to a vocal input. The comparison between the algorithms was made taking into account the mean-square error (MSE) and average attenuation (AV), as well as the required number of multiplications.

Sub-band adaptive filtering for acoustic echo cancellation

Sub-band filtering is employed in various system identification applications. It reduces the computational complexity and adaptation time. The paper focuses on sub-band filtering for acoustic echo cancellation. A Simulink model for 2, 4, 8 sub-bands was created and simulated. The echo is created by applying a vocal test to a first Schroeder reverberator. The performance of the full- and sub-band echo canceller is measured in terms of error signal and echo-return loss enhancement.

Nonlinear acoustic system identification using a combination of Volterra and power filters

The paper proposes a nonlinear system identification method that uses a combination of adaptive linear, Volterra and power filters. Adaptation of the kernels is made using a Normalized Least Mean Square algorithm. The method is applied in echo cancellation, where several sources of nonlinearities exist: the overdriven amplifier, the small loudspeaker at high volume, the room with different absorbent walls. Functions with nonlinear characteristics are chosen to model these distortions. The evaluation is made in terms of Echo Return Loss Enhancement. Results show that the overall convex combination approach performs better or at least as well as the best single adaptive filter.

Image rejection filter based on blind source separation for low-IF receivers

In this paper the low-IF receiver architecture, the image interference due to the IF domain, phenomena also referred as I/Q imbalance and a modification for a blind source separation algorithm for image rejection are presented. The modified algorithm considers all the received samples of IF signal for the estimation of the I/Q imbalance parameters. The resulting image rejection filters major advantage is the single clock domain implementation at the cost of power consumption increase. A Simulink model was created to check the availability of the enhanced algorithm, its performance measured and compared to the existing implementation. Simulations show that the proposed filter achieves the same performance in terms of IRR as the reported method.