WO1985003359A1

WO1985003359A1 - Method of processing sensor elements

Info

Publication number: WO1985003359A1
Application number: PCT/AU1985/000007
Authority: WO
Inventors: Antonio Cantoni
Original assignee: The Commonwealth Of Australia Care Of The Secretar; The University Of Newcastle
Priority date: 1984-01-23
Filing date: 1985-01-16
Publication date: 1985-08-01
Also published as: NO853629L; JPS61501950A; EP0168442A1

Abstract

A method of improving the reception of broadband and narrowband signals by optimally processing the outputs of an array of sensor elements, either electromagnetic or acoustic, in which outputs are passed through channel splitters (block 1) and then a series of time delays (T) and weighting coefficients (W) before finally being summed, the weighting coefficients (W) being adaptively updated by using a set of quadratic constraints imposed on the weighting coefficients so that noise from other directions is optimally rejected.

Description

METHOD OF PROCESSING SENSOR ELEMENTS

This invention relates to a method of processing sensor elements and applies to the processing of both electromagnetic fields or sound fields.

5- The method of the invention is based on the solution of a new optimisation problem devised to ensure minimal suppression of signals within a specified angle of arrival while maximising the rejection of other undesired directional and non- 10, directional signals. The technique also ensures that the overall processor maintains a frequency response in the look direction which is within a specified tolerance of a desired frequency response.

The technique of the invention is applicable 15. to both broadband and narrowband situations, and is applicable to the following pre-steering situations:

(a) exactly pre-steered

(b) coarsely pre-steered (inexact/discretized)

(c) without pre-steering.

20. The technique of this invention also permits various types of errors and uncertainties about parameters to be included in the processing, thereby reducing the sensitivity of the performance of the processing to these errors and uncertainties,

25. and can be applied to a single fixed look direction, many fixed look directions, or can be used to scan continually.

The basis of the invention is the use of series of arrays of broadband time delay and weighting 30. coefficient units arranged as multiple control means in a series of channels. To fully understand the scope of the invention, reference may be had to the specification of United States Patent No. 3,852,707 of Sameul W. Autrey assigned to the United States of America as represent- 5. ed by the Secretary of the Navy which shows a process¬ or which provides a constant beam width radiation pattern over an extended range of frequencies. In this weights are used, but they are fixed by the frequency range and do not adapt to optimally suppress 10. interfering noise from other directions.

The specification of United States Patent No. 4,060,792 of Arent van Heyningen, assigned to the Raytheon Company describes a processor having a set of fixed weights to produce a given sidelo_^be level for 15. clipped signals but these are not adjusted to suppress noise.

The specification of United States Patent No. 3,766,559 of Wade E. Butcher and Robert J. Sims assigned to Harris-Intertype Corporation has signals 20. passed through tapped delay lines which are weighted and then summed, and use a pilot signal.

The basis of the applicants invention is the basis of calculating the weights by using the summed output and a quadratic constraint. The quadratic 25. constraint obviates the need for a pilot signal to maintain the desired look direction response.

The present invention thus comprises a method of processing electromagnetic or acoustic signals in any medium to improve the signal in broadband and narrow- 30. band signal processing, using an array of sensor elements. The sensor outputs are first optionally passed through channel splitters and then a series of time delays and weighting coefficients before finally being summed. The weighting coefficients are adaptively „ 5. updated in such a way that noise from other directions is optimally rejected. A set of quadratic constraints is imposed on the weighting coefficients to give tele¬ processor robust behaviour to errors arising in the system.

10. The invention can be realised computationally in a number of ways but to enable it to be readily appreciated, the description will be made with reference to the accompanying drawings in which:

FIG. 1 is a block diagram showing how an array 15. of sensors passes signals to a pre-processor and this passes the partly processed signal to the main processor which provides the output signal,

FIG. 2 is a block diagram to show how the signals from block 1 can be processed to provide 20. enhanced processed output. The sensor outputs are passed through signal splitters and then a series of time delays and weighting coefficients before being summed,

FIG. 3 is a block diagram similar to FIG. 2 25. but omitting the signal splitters,

FIG. 4 is a definition of the symbols used, and

FIGS. 5(a) and 5(b) show the principle of the processor.

In FIG. 5(a) the weight vector w is adjusted to

30. minimise the interference power whilst keeping look-direction response within a certain mean-square distance of the desired response. FIG. 5(b) shows how this adjustment is carried out - to keep the integrated square deviation in response δ w must remain within or on the hyperellips- oid "A". Setting __w to the point "E" minimises the 5. output power (and hence the interference power) whilst remaining on the hyperellipsoid "A".

The signals are generated by an arbitrary array of sensors...distributed in space, whose sensors feed processing elements block 1 and block 2.

10. The sensors may be designed to respond to any type of signal propagating in the medium in which the array is immersed or with which the array is in contact, and as stated could be designed to sense electromagnetic fields or sound fields.

15. The medium in which the fields are propagating could be a solid, a liquid, a gas or a vacuum depend¬ ing on the nature of the field.

The design of the block, labelled main processor, is such that an infinite variety of processing

20. elements in the block 1 can be catered for. For example, block 1 may not exist at all. Another possibility is that the block 1 may consist of devices which delay the array sensor signals in such a way as to realign a wavefront incident on

25. the array from a specific direction by compensating exactly for the array geometry. This is referred to as exact pre—steering. Another possibility is that the block 1 consists of devices which only approximately compensate for a specific wavefront.

30. Yet another possibility is that block 1 has some arbitrary combination of filters and time delays. Block 2 is the main processor block which consists of a network which enables the L signals received from Block 1 to be combined in a linear manner with weighting defined by coefficients w.. ^■ 5. Examples of some possible weighting networks are illustrated in FIG. 2 and FIG. 3.

A technique for specifying the weighting co¬ efficients of the main processor block has been developed, such that irrespective of the array

10. geometry and the contents of Block 1 the overall system enables unwanted signals to be rejected while maintaining a desired response to a specific wavefront. For example, the desired response may be a flat frequency response for a plane wave incident on

15. . he array from a specific direction.

The weighting coefficients can be determined adaptively- or can be fixed and are determined by solving the following optimization problem.

minimize P(W_) (1)

20. subject to E(W) < 6 (2)

(a) P(.W) is a measure of the total power at the output of the processor and is a function of the weighting coefficient vector W_. For example P(,W) could be the ensemble average power for an 25. assumed signal environment, in which case P(W_) would be a quadratic function of the form

P(W_) = W^TRW (3) where R is the correlation matrix for the processor. P(W) could be an estimate of the output power of the processor in which case

P(W) = W%_

5. where "R. is an estimate of the correlation matrix for the processor. __

(b) E(_W_) is a measure of the deviation of the response of the processor, from some desired response, to a specific wavefront. For example,

10. it can be shown that for a mean square type of measure of deviation and for linear weighting in block 2 E(W.) has the form

E(W.) = W^TQW - 2P^TW + d

where Q, P and d are constant parameters depending on 15. the specific application.

Thus in this case, inequality 2 becomes a quadratic inequality constraint,

(c) δ is a parameter which determines how closely the overall processor achieves the desired

20. response.

Claims

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:

^" 1. The method of improving the reception of broadband and narrowband signals by optimally process¬ ing the outputs of an array of sensor elements, either 5. electromagnetic or acoustic, disposed in any medium, with or without pre-processing the sensor, outputs being first optionally passed through channel split¬ ters and then a series of time delays and weighting coefficients before finally being summed, the weight-

10. ing coefficients being adaptively updated so that noise from other directions is optimally rejected, characterised in that the processor uses a set of quadratic constraints imposed on the weighting coefficients whereby robust behaviour to errors

15. ' arising in the system results.

2. The method of claim 1 in which the weighing coefficient is determined adaptively or is fixed and determined by solving the following optimisation problem

20. minimize P(W) subject to E(W) <. s

where P(W) is a measure of the total power at the output of the processor and is a function of the weighting coefficient vector W and E(W) is a measure 25. of the deviation of the response of the processor, from some desired response, to a specific wavefront.

3. The method of- claim 2 wherein P(W) is the ensemble average power for an assumed signal environ¬ ment, in which case P(W) would be a quadratic function

30. of the form

P(W) = W^TRW where R is the correlation matrix of the processor. P(W) could be an estimate of the output power of the processor in which case

P(W) =- W%f

5. where R is an estimate of the correlation matrix for the processor.

4. The method of claim 2 wherein the mean square type of measure of deviation and for linear weighting in block 2 E(W) has the form

10. E(W) -= W^TQW - 2P^TW + d

whereby the constraint in claim 2 becomes quadratic inequality constraint, with Q, P, and d being constant parameters depending on the specific application.