DIRECTIONAL NOISE AMBIGUITY RESOLUTION WITH LINEAR ARRAYS* 



Gordon Raisbeck 



Arthur D. Little, Inc. 



Cambridge, Massachusetts 02140 



(Received 8 October 1974) 



ABSTRACT 



Often the most convenient tool with which to measure the azimuthal distribution 

 of underwater noise is a horizontal linear array. Unfortunately, a linear array cannot dis- 

 tinguish right from left, and measurements made with it are ambiguous. The ambiguities 

 can be reduced by making measurements with the array pointed in two or more direc- 

 tions. The choice of number and alignment of array directions and of method of reduc- 

 ing the data depends on the criteria for v/eighting errors, which in turn depends on the 

 use to be made of the results. No one scheme is optimal for all applications. The merits 

 and weaknesses of several which have been used are compared. 



INTRODUCTION 



In surveillance sind many other practical detection problems, phenomena are differentiated a priori 

 into interesting events, which we may call targets, and uninteresting events, which we may call noise and 

 interference. The technical interest in the detection problem comes from the fact that the information 

 available to the decision maker only partly specifies the type of event. Given only incomplete informa- 

 tion, we must judge as reliably as we can whether the event was a target or not. 



It often happens that most of the packets of information, which we shall call signals, that we re- 

 ceive about target events share a common chsiracteristic which is absent from the signjds from nontarget 

 events. One effective detection tool is to sort or match signals according to the presence of absence of 

 such characteristics. To apply this technique, we must know something about the characteristics of sig- 

 nals from target events and from nontarget events. It is not sufficient to have information on particular 

 signals: we must have some information about the range and distributions of all possible signals, and 

 direct our attention to those cheiracteristics which effectively discriminate between targets and nontargets. 



No ship or submarine can operate at sea without converting some of its mechanical energy into 

 low-frequency acoustic energy. Furthermore, acoustic energy is propagated long distances in the ocean 

 with only moderate loss. Therefore, the unavoidable acoustic emissions from ships and submarines can 

 be measured far away. Unfortunately, the ocean is quite noisy. Therefore, the primjiry problem in pas- 

 sive acoustic submarine detection and classification is not one of making a sufficiently sensitive acoustic 

 measuring instrument, but rather one of distinguishing an amply loud acoustic signal from a submcirine 

 from the even louder ambient noises and interference. 



One of the ways in which signals aie distinguished from noise is by their spatial distribution. Under 

 most conditions, most of the acoustic energy from a tetrget will arrive at a receiver along either one or 

 only a few propagation paths, presenting wavefronts at the receiver which are locally nearly plane, 

 ambient noise, on the other hand, has a vastly different spatial distribution. If the acoustic energy at 

 the receiver is represented as the sum of incoming pljine waves (and with some restrictions of little 



♦This paper is based on a presentation of the same title delivered at the LRAPP Review of Mid-Water Acoustic Measurement 

 Systems at the Woods Hole Oceanographic Institution on 7 June 1974. 



889 



