A Derived Robust Statistics Approach For Adaptive Volterra Filters Applied In Nonlinear Acoustic Echo Cancellation Scenarios
Journal Title: Acta Technica Napocensis- Electronica-Telecomunicatii (Electronics and Telecommunications) - Year 2013, Vol 54, Issue 1
Abstract
The paper proposes a novel updating concept for adaptive Volterra kernels that relies on a robust statistics approach. The optimization of a certain cost function leads to the update equations for the linear and quadratic kernels with respect to an optimum error threshold, set according to the relative local noise power. Thereby, for absolute error samples larger than the threshold, which occur mainly in the convergence stage of the adaptive filter, the adaptation is achieved using a sign version of the Normalized Least-Mean-Square algorithm. In the saturation stage of the filter, when most of the absolute values of the error samples are smaller than the particular threshold, the adaptation is carried out based on the Normalized Least-Mean-Square algorithm. This technique tends to eliminate the influence of large valued outliers by improving the convergence of the conventional Volterra filters and maintaining the same misadjustment in acoustic echo cancellation setups. The acoustic system is designed using measured Volterra kernels. The method is tested in terms of Echo Return Loss Enhancement for different probability density functions of the source signal. Also, for the white Gaussian noise case as excitation, a new convergence attribute is introduced for a more precise comparison of the adaptive methods.
Authors and Affiliations
Cristian CONŢAN, Marina Dana ŢOPA, Botond Sandor KIREI, Ingrid Maria KOVACS
Keywords
Related Articles
- EP ID EP125611
- DOI -
- Views 76
- Downloads 0
A PSPICE Model For T Matching Circuits
In electronics, impedance matching is an engineering technique employed in circuit design for matching unequal source and load impedances. This can be done by inserting a matching circuit between the source and a load. F...
A Functional Model for A Standalone Fingerprint Recognition System
Abstract: The paper describes a functional model for a standalone fingerprint recognition system, representing a minimum implementation to demonstrate the functionality of such system. For simplicity, the model uses only...
Retrieving Call Detail Records from Asterisk using SNMP
Several SNMP agents could be integrated by IETF’s AgentX protocol which separates the communication with SNMP manager in two parts: AgentX between sub-agents and master agent, and SNMP between master agent and manager. T...
Simulation of the Ford–Fulkerson Algorithm Using Omnet++
This paper presents a contribution to the OMNeT++ 4.1 simulator in C++ by integrating our own implementation of the FF - Ford-Fulkerson multipath routing algorithm. In order to validate it, an eight-node testbed works wi...
Study of Effects of The Short Time Fourier Transform Configuration On EEG Spectral Estimates
EEG signals recorded from scalp contain useful information about the activity of a large number of neurons. Signal processing is needed to extract this information from the EEG signal. Here we study the effects of config...