Sven Nordholm

A spatial filtering approach to robust adaptive beaming

The problem of controlling the superresolution in adaptive beamformers is treated. A straightforward method is presented that works for both narrowband and broadband arrays. The method is based on forming the blocking matrix in a general sidelobe canceller structure using a spatial FIR filter. The suppression of this spatial filter and the implicit noise of the leaky least-mean-square algorithm together determine the beamformer. >

The broad-band Wiener solution for Griffiths-Jim beamformers

The authors treat the problem of finding the Wiener filters for a wideband adaptive beamformer. Explicit expressions are given for the filters and error power spectrum in the frequency domain. The expressions are simple to program and give the possibility of investigating superresolution for wideband signals and of determining sufficient lengths of the finite-impulse-response filters in adaptive arrays. The major results are an analytical expression for the broadband array Wiener solution given for an arbitrary spectrum for target, interference, and noise and an approximate analytic expression for the angular 3-dB beamwidth for broadband signals. >

A multi-DSP implementation of a broad-band adaptive beamformer for use in a hands-free mobile radio telephone

An implementation of a broadband adaptive array on a multiprocessor digital signal processing system for use in a hands free mobile radio telephone is described. This implementation of a five-microphone adaptive Griffiths-Jim array can handle FIR filters with up to 128 taps behind each microphone at a sampling rate of 8 kHz. The filter structure makes it possible to combine an adaptive array with a noise canceler. The near-field problem has been solved by using focusing, a speech-controlled adaptive algorithm, and a short hourglass. Preliminary measurements indicate a considerable potential for this technique in hands-free mobile telephony. The array gives a 20-30 dB suppression of a broadband jammer covering 300-1100 Hz, even with three reflecting walls surrounding the microphone. >

Noise cancelling convergence rates for the LMS algorithm

Abstract This paper treats the convergence characteristics for the LMS algorithm combined with a conventional FIR filter in two applications, identification and equalisation. It is well known that the convergence rate is connected to the eigenvalue spread, but often no distinction is made between convergence rate is connected to the eigenvalue spread, but often no squared error, i.e. the learning curve. It is, however, shown that the eigenvalue spread can sometimes be a misleading quantity when judging convergence rates. We will show that the eigenvalue spread is less significant when working with identification in the form of noise cancelling, while it is of vital importance in inverse filtering problems such as equalisation.

Chebyshev Optimization of Circular Arrays

The lobe design of a narrow-band antenna array is formulated as a complex Chebyshev optimization problem. Previously, the solution of similar problems have been approximated using various techniques including conventional finite-dimensional linear programming. In this paper, recently developed semi-infinite linear programming techniques employing simplex extension algorithms are used to design the required antenna array.

Superresolution effects in broadband antenna arrays

The problem of finding the angular 3-dB beamwidth for a wideband Griffiths-Jim adaptive beamformer is discussed. Approximate analytic expressions are presented and their relation to the Cramer-Rao bound (CRB) is considered. Results on the CRB in direction of arrival estimation for a sinusoidal signal are extended to narrowband and broadband signals, together with simple interpretations of their behavior. Superresolution results with a five-element array for three target signal bandwidths (400-2000, 800-1600, and 1120-1280 Hz) are presented.<<ETX>>