Improved finite samples resampling for unbiased wide-band direction of arrival estimation

Abstract

This paper considers wide-band DOA estimation using the spatial resampling method for coherent averaging. In this paper a rough estimate of the resampling error distribution is obtained. Careful examination of this distribution yields an improved resampling formula suitable for, but not limited to, arrays with finite sensors. It is shown that the proposed formula yields resampling error much less than that of conventional resampling. Moreover it is shown how the resampling error can be more reduced by proper choice of the focusing frequency. Based on the this resampling formula a perfect focusing scheme is then proposed and its performance is evaluated and compared to other wide-band methods through simulations. Simulations showed quite satisfactory and robust performance of the proposed scheme. It was shown that it succeeds in situations where all considered wideband methods fail. In addition it is bias-free and can be implemented quite efficiently. Hence it is quite suitable for reliable real-time DOA estimation in reverberant environments.

Efficient array design algorithm for wide-band application of the MUltiple SIgnal Classification algorithm

“…In addition, using the approximate Eq. (B•9) for calculating the virtual array output, P v m ð f Þ, does add a considerable amount of error [12]. Due to these effects the ideal model of Eq.…”

“…The concept of coherent averaging was first introduced by Wang and Kaveh in [7]. Since then a number of different methods following the same concept were introduced as in [8][9][10][11][12]. In the following coherent averaging by the spatial resampling method [8] will be considered as an example of coherent averaging methods.…”

“…Reverberation not only yields the problem of coherent sources but also increases the number of sources to be resolved. On the other hand, wide-band nature of sound signals led to the development of some wide-band generalizations as incoherent averaging [6], coherent averaging [7] and focusing techniques [8][9][10][11][12]. Most of these methods can also handle the coherent sources problem and rely on applying eigenvector based methods on an array whose sensor spacing is considerably less than half wave length.…”