P020

A New Simultaneous Estimation of Directions of Arrival and Spectral

Copmponents

A0053 Peyman P. Moghaddam Earthquake Research Institute, Unversity of Tokyo

A0054 Junzo Kasahara Earthquake Research Institute, Unversity of Tokyo

In this paper, we develop a new method to estimate the Directions of

Arrival (DOA)of some possibly coherent wideband sources impinging on a

uniform linear array. This method is based on Least Square Error

criterion (LSE). A comparison between this method and other previous

methods such as Spatial Smoothing and Root-MUSIC is made and the

results show that the new estimator had superior performance than

other methods. We also show that mentioned method could

simultaneously estimate the spectral components of each source, which

can generally be used as a classifier.

The problem of better separation (angle estimation and classification)

of signals impinging to an array under different circumstances is

considered. In recent decade, huge researches conducted in this area

in many applications such as underwater acoustics and mobile

communications.

Direction of arrival estimation is one of the most important parts of

any array system. When some signals arriving from different angles of

arrival impinging on an array of sensors the main purpose is to

estimation these angles. The success of high-resolution eigen-value

decomposition techniques such as MUSIC, ESPIRIT or any adaptive

beamformers depends on the severity of correlation among sources.

Poor resolution is to be expected, as correlation among sources is to

increase. Two versions of high-resolution techniques that deal with

multipath scenario are forward only and forward/backward spatial

smoothing (FBSS). These are preprocessing methods for decorrelating

signals. They form a spatially smoothed covariance matrix by grouping

an equally spaced array into subarrays and then average the spatial

covariance, and subsequently, any eigen-based methods can successfully

be applied for decorrelating signals regardless of coherence among

them.

Estimation of all the parameters of sources, which impinging to an

array, required solving a Maximum Likelihood (ML) function with large

number of variables, which is impossible. This reason caused to

finding some simple solution of ML in different circumstances. For

example, an approach is based on large sample maximum likelihood

algorithm to estimate the DOAs and amplitudes of known coherent

signals arriving at an array of sensors. This can further be

complicated if the transmitted signal have gone through a multipath

channel. One of the important abilities of array processing is to

detect the spectral characteristics of signals impinging on an array.

The new method introduced in this paper provides simultaneous

estimation of the DOAs and spectral components of signals that are

coherent, possibly due to a multipath channel. Previously introduced

methods eventually implement some version of spatial smoothing that

are eigen-based. The new method is not eigen-based rather uses LSE

criterion on the received data into array. Despite of eigen-based

methods, this method required to solve a multi-dimensional minimizing

function which its dimension equal to number of source angles.

At the end, we show that the new method has better performance than

the spatial smoothing to estimate the parameters of some possibly

wide-band unknown signal in a multipath environment.