Closed-loop multistage beamformer

Details Closed-loop multistage beamformer Closed-loop multistage beamformer

2000-10-20

2003-07-22

Barlow, John (Department: 2863)

Data processing: measuring, calibrating, or testing

Measurement system

Measured signal processing

C702S189000, C702S190000, C702S191000, C342S378000, C375S148000

Reexamination Certificate

active

06598014

ABSTRACT:

BACKGROUND OF THE INVENTION
The invention relates to adaptive beamforming, and in particular to a closed-loop multistage beamformer.
In this application, the following notation will be used. Bold upper and lower case letters will denote matrices and vectors, respectively. Scalars are italicized. Superscript H denotes conjugate transposition. Expectation is denoted by E{·}. Other nomenclature is summarized in Table 1
To avoid confusion, Table 2 defines some terminology as it will be used in this application. Applicants are careful to point out the duality between spatial array processing and other related applications. The subsequent presentation will consider only multi-sensor (spatial) array processing; however, it should be understood that the principals of the architecture could easily be employed in these other applications as well.
In general, sensor arrays are capable of rejecting undesirable interference and recovering signals of interest. When the characteristics of the interference are unknown, smart signal processing techniques can be used to “adapt” to the interference. This information can then be used to reject the undesired interference. This type of processing is broadly termed “Adaptive Beam Forming” (ABF). For over 30 years, different “smart sensor” algorithms (and the beamforming hardware required to implement them) have been developed to accomplish this goal. To evaluate these techniques, one typically measures their performance, i.e., their ability to recover signals and suppress interference. One also considers the cost of building suitable array hardware (e.g., receivers and I/O links) and signal processing hardware (e.g., nonadaptive and adaptive filters) as required to implement these algorithms. The goal, of course, is to select techniques that attain a very high level of interference suppression at a low cost.
Oftentimes, sensor arrays are designed to have large apertures, obtaining very high sensitivity levels. Such arrays may be highly digitized (either at the element or subarray levels) to improve performance (e.g., by increasing the dynamic range and/or improving system sensitivity and interference rejection). This high level of digitization results in many “degrees of freedom” (DOFs) that can be used for adaptive beamforming. However, the processing hardware required to exploit these DOFs can be prohibitive.
DOFs can be exploited in a fixed fashion, an adaptive fashion, an instantaneously adaptive fashion, or some combination thereof. Applicants associate the type of DOF exploitation with the degree of flexibility possible within an array processing system at design time. A fixed DOF refers to a DOF that can not be adjusted by the processor in a real-time data-dependent fashion. Fixed DOFs are thus maximally inflexible. An adaptive DOF, as defined here, refers to a DOF that can be under real-time processor control, i.e., it can be controlled in a data-dependent fashion. Adaptive DOFs are thus somewhat flexible. An instantaneously adaptive DOF, on the other hand, refers to an adaptive DOF that is under the direct control of the processor at some given instant. Instantaneously adaptive DOFs are thus maximally flexible. Note that this latter term is introduced specifically to distinguish the operation of the invention from that of other ABFs.
To clarify these methods of DOF exploitation, an analogy is offered. A sensor array is analogous to a black box with a large number of control knobs on the outside. A DOF can be thought of as a single control knob. An operator (the processor) monitors the operation of the black box, and adjusts control knobs as needed. A fixed DOF can be thought of as a control knob that the operator is not permitted to touch. An adaptive DOF can be thought of as a control knob that an operator could move. An instantaneously adaptive DOF can be thought of as a control knob that is actually being manipulated (by the operator) at some given instant. Note that while the number of control knobs may be large, the operator may not be capable of turning them all at once. Hence the number of instantaneously adaptive DOFs may be much lower than the number of adaptive DOFs.
Next, the adaptive beamforming problem can be succinctly stated as follows. Suppose one has a sensor array that produces N digitized channels. As such, there are a total of N DOFs available. A sample vector (a.k.a. snapshot) associated with processing time k is given by the N×1 column vector, x
k
. The matrix X
k
denotes a N×L set of snapshots at this time. The interpretation of x
k
and X
k
will depend on the specific array processing application. For example, in some radar fields x
k
would be a single snapshot from CPI k, and X
k
would be the set of all snapshots from CPI k. These snapshots contain energy from J jammers and noise. The task is to build a filter capable of suppressing this interference while receiving signals from direction &THgr;.
There are currently two general classes of ABF techniques that might be used to solve this problem. The first general class of ABF techniques is called Fully Adaptive Beam Forming (Full ABF). Full ABF methods use snapshots from all N DOFs (simultaneously) to adapt to the jamming in this N dimensional space. In this sense, Full ABF works with N instantaneously adaptive DOFs. For representative techniques, see B. Van Veen, “Beamforming: a versatile approach to spatial filtering,” IEEE ASSP Magazine, April 1988, section IV-V, incorporated herein by reference.
Full ABF thus requires the hardware to simultaneously collect and process data from all N channels. For many algorithms in this class, processing complexity grows as N
3
. Thus, for large digitized arrays (i.e., large N), the processing complexity and associated size, weight and power can be quite large. Furthermore, the convergence time of the adaptive algorithm typically increases proportional to N. Thus, for large, highly digitized arrays, a large training interval will be required (this is also quite undesirable). Furthermore, adaptive beamforming is typically proceeded by other processing (e.g., filtering) which might then be performed on all N channels, adding to complexity. Lastly, the adaptive processor itself is often situated in a location that is separated from the array (e.g., below deck on a ship). Thus, Full ABF requires a communication network with enough bandwidth to carry all N channels of the array data to the processor.
With sufficient training data, Full ABF techniques will achieve near optimal performance—albeit at a high cost. Problems occur, however, when sufficient training data is not available, or costs are constrained.
The second general class of ABF techniques attempts to achieve good performance (in restricted environments) and rapid convergence at a greatly reduced cost. These techniques, collectively known as Beamspace Adaptive Beam Forming (Beamspace ABF), achieve this goal by accepting poorer performance in environments with a large number of jammers.
Beamspace ABF begins by mapping the N array channels into N′ beams prior to adaptive interference rejection, as shown in FIG.
1
. In this sense, Beamspace ABF works with only N′ instantaneously adaptive DOFs (the remaining N−N′ DOFs are all fixed DOFs). This mapping is determined by a N×N′ linear transformation matrix, T, which is called the “beamspace transformation matrix.” The beamspace transformation is used to create a set of N′ ×1 beamspace snapshot vectors, Y
k
, via:
Y
k
=T
H
X
k
  (1)
(a single such beamspace vector is denoted y
k
).
Next, adaptive processing is performed with the goal of adapting to the jamming. This processor has access only to the beamspace snapshots. Thus, the adaptation can be viewed as taking place within a N′ dimensional space. For representative techniques, see B. Van Veen, “Beamforming: a versatile approach to spatial filtering,” IEEE ASSP Magazine, April 1988, section VI, incorporated herein by reference. At this stage, the adaptation methods that are employ

Affiliated with

Also associated with

Rating

Closed-loop multistage beamformer does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Closed-loop multistage beamformer, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Closed-loop multistage beamformer will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3103918