Data processing: artificial intelligence – Neural network – Structure
Reexamination Certificate
1999-08-10
2003-02-11
Davis, George B. (Department: 2121)
Data processing: artificial intelligence
Neural network
Structure
C706S042000
Reexamination Certificate
active
06519577
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to apparatus and methods for digital filtering of signals, for example to remove shot or impulse of noise on analogue or digital signals.
2. Discussion of Prior Art
Digital filters can be used in many different applications on a wide variety of signals and structures. For example the signal structure may be 1-dimensional, e.g. audio signals, instrumentation signals, etc. It may be 2-dimensional e.g. a video image, a RADAR or sonar image, etc. It may be 3-dimensional such as a representation of a volume, e.g. a volume of air space for Air Traffic Control, or a scanned volume in a 3-dimensional medical imaging technique and so on. Mathematically this technique is applicable to signals of any dimensionality.
SUMMARY OF THE INVENTION
Accordingly, in one aspect, this invention provides a digital filter for filtering a digital signal comprising digital words which comprises:
window defining means for defining a window containing a plurality of digital words around a word to be filtered;
means for supplying bits from corresponding bit positions of words in said window to respective sum and threshold means;
said sum and threshold means being operable to output a preset output bit if the sum of the bits applied to said sum and threshold means passes a preset threshold, and
output means for outputting a digital word corresponding to a generic result or filter output represented by the outputs of said sum and threshold means.
The digital data represents a 1-, 2-, 3- or higher dimensional signal structure.
Preferably said means for supplying supplies said bits to a single group of sum and threshold means, thereby to output a serial stream of digital words corresponding to generic outputs.
In a parallel fed arrangement, said means for supplying may supply said bits to a plurality of groups of sum and threshold means, thereby to output a parallel stream of digital words corresponding to generic outputs.
The threshold applied by each of said sum and threshold means may be uniform, or it may be non-uniform or adjustable.
Each of said sum and threshold means may comprise a hierarchy of sum and threshold devices in a layered structure, for example as described in U.S. Pat. No. 6,272,511.
Each of said sum and threshold means may comprise a Hamming value comparator made up of a plurality of bit manipulation cells.
Where required, the digital signal filter may comprise unit Hamming distance code converter means for converting said input digital words into unit Hamming distance code prior to supply to said sum and threshold means, and decoder means for returning the filtered unit Hamming distance code to binary.
Where the digital signal filter is required to filter an analogue signal, it preferably comprises analogue to digital conversion means for converting said analogue signal to digital words for filtering, and digital to analogue conversion means for returning said filtered digital words to an analogue signal after filtering.
Preferably said means for supplying supplies to said sum and threshold means bits from corresponding bit positions of words in said window, including said word to be filtered. Alternatively said means for supplying may supply to said sum and threshold means bits from corresponding bit positions in said window, excluding said word to be filtered.
Preferably, said digital filter includes means for comparing said generic result with said data representing said word to be filtered and for selecting for output said generic result or said word to be filtered on the basis of said comparison. Additionally, said means for comparing may further comprise correlation means for comparing the bits set at each bit position in said generic result and said data representing said word to be filtered, and further sum and threshold means for summing and thresholding the results of said correlation.
Whilst the invention has been described above, it extends to any inventive combination of features set out above or in the following description.
REFERENCES:
patent: 4674125 (1987-06-01), Carlson et al.
patent: 4805129 (1989-02-01), David
patent: 4868733 (1989-09-01), Fujisawa et al.
patent: 4868773 (1989-09-01), Coyle et al.
patent: 5029305 (1991-07-01), Richardson
patent: 5063521 (1991-11-01), Peterson et al.
patent: 5113507 (1992-05-01), Jaeckel
patent: 5253330 (1993-10-01), Bamacher et al.
patent: 5382955 (1995-01-01), Knierim
patent: 5459466 (1995-10-01), Knierim et al.
patent: 5854636 (1998-12-01), Watanabe et al.
patent: 5892962 (1999-04-01), Cloutier
patent: 5940529 (1999-08-01), Buckley
patent: 5974521 (1999-10-01), Akerib
patent: 6185336 (2001-02-01), Clark et al.
patent: 6205197 (2001-03-01), Ananth
patent: 0 217 009 (1987-04-01), None
patent: 0 221 238 (1987-05-01), None
patent: 0 230 752 (1987-08-01), None
patent: 2 095 067 (1982-09-01), None
patent: 2 223 369 (1990-04-01), None
patent: 5-243911 (1993-09-01), None
patent: 5-243912 (1993-09-01), None
patent: 7-95091 (1995-04-01), None
patent: 7-160796 (1995-06-01), None
patent: 7-109988 (1995-11-01), None
patent: 8-505741 (1996-06-01), None
patent: 2605265 (1997-02-01), None
patent: 2629748 (1997-04-01), None
patent: 94/16499 (1994-07-01), None
patent: WO 94/16499 (1994-07-01), None
Kennedy et al, “C-NNAp-a parallel Processing Architecture for Binary Neural Networks”, IEEE International Conference on Neural Networks, Nov.-Dec. 1995.*
Kim, et al, “Memory Based Processor Array for Artificial Neural Networks”, IEEE International Conference on Neural Networks, Jun. 1997.
BAE Systems plc
Davis George B.
Nixon & Vanderhye P.C.
LandOfFree
Digital signal filter using weightless neural techniques does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Digital signal filter using weightless neural techniques, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Digital signal filter using weightless neural techniques will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3142894