Loading [MathJax]/extensions/MathZoom.js
Minimax design of nonnegative finite impulse response filters | IEEE Conference Publication | IEEE Xplore

Minimax design of nonnegative finite impulse response filters


Abstract:

Nonnegative impulse response (NNIR) filters have found many applications in signal processing and information fusion areas. Evidence filtering is one of the examples amon...Show More

Abstract:

Nonnegative impulse response (NNIR) filters have found many applications in signal processing and information fusion areas. Evidence filtering is one of the examples among others. An evidence filter is required to satisfy a nonnegativity condition and a normalization condition on its impulse response coefficients, and thus is basically an NNIR filter. This paper considers the design of nonnegative finite impulse response (FIR) filters based on frequency response approximation and proposes a constrained minimax design formulation using the fundamental limitations on the NNIR filter's frequency responses recently developed in the literature. The formulation is converted into a linearly constrained positive-definite quadratic programming and then solved with the Goldfarb-Idnani algorithm. The proposed method is applicable to nonnegative FIR lowpass as well as other types of filters. Design examples demonstrate the effectiveness of the proposed method.
Date of Conference: 09-12 July 2012
Date Added to IEEE Xplore: 30 August 2012
ISBN Information:
Conference Location: Singapore

I. Introduction

Distributed sensor networks consisting of a large number of low-cost and low-power sensors have found many applications in military surveillance, homeland security, traffic surveillance, environment monitoring, etc., for collecting observations and processing information. The use of distributed nodes with multiple sensing modalities can significantly enhance the robustness and the accuracy of the decision making process in such environments [1]. Data collected by nodes are fused at various levels and in different ways in order to make useful inferences at the decision-making center. The inferences made based on the fused information range from distinguishing among different event (or threat) types, estimating important parameters such as location or velocity of an object, target tracking, as well as detecting the presence or absence of events with certain “frequency” or “spectral” characteristics [2].

References

References is not available for this document.