Fast pattern recognizer utilizing dispersive delay line

Claim 2

A pattern recognition system which compares an unknown pattern with a known pattern, the system comprising: A. a first processing means for transforming said unknown pattern to the frequency domain, thereby providing a processed unknown pattern; B. a second processing means for transforming said known pattern to the frequency domain, thereby providing a processed known pattern; C. a multiplying means for multiplying a chirp signal f ch , the processed pattern produced by one of said processing means, and the complex conjugate of the processed pattern produced by the other of said processing means, thereby providing a product signal; D. a dispersive delay line; and E. a means for applying said product signal to said dispersive delay line, thereby providing a correlation signal indicative of the level of correlation between said unknown pattern and said known pattern.

Claim 3

The pattern recognition system of claim 1 wherein said dispersive delay line is a reflective array compressor.

Claim 4

The pattern recognition system of claim 1 wherein each of said means for transforming comprises means for performing a Fourier transform.

Claim 5

The pattern recognition system of claim 1 further comprising: A. a peak detector which receives said correlation signal and detects the peak of said correlation signal over a certain time period, and provides a peak detected signal indicative thereof; and B. an analog-to-digital converter which receives and digitizes said peak detected signal, and provides a digitized peak detected signal indicative thereof.

Claim 6

The pattern recognition system of claim 1 wherein said dispersive delay line comprises a single input transducer which receives said modulated signal and a single output transducer which provides said correlation signal.

Claim 7

The pattern recognition system of claim 1 wherein said correlation signal also provides an indication of the spatial region in said unknown pattern which correlated with said known pattern.

Claim 8

The pattern recognition system of claim 1 further comprising means for generating said known pattern which comprises an analog-to-digital converter which digitizes a pattern signal of bit-mapped data thereby providing a known pattern.

Claim 9

The pattern recognition system of claim 1 further comprising means for dividing said unknown pattern into a plurality of unknown sub-patterns and sequentially applying each of said plurality of unknown sub-patterns to said first processing means, said first processing means providing a plurality of processed unknown sub-patterns, said system further providing a plurality of correlation signals each of which is uniquely associated with the level of correlation between said known pattern and one of said unknown sub-patterns.

Claim 10

The pattern recognition system of claim 8 wherein each of a plurality of resulting processed unknown sub-patterns is a sub-array of equal size, and wherein the resulting product signal is continuously fed into said dispersive delay line.

Claim 11

The pattern recognition system of claim 1 further comprising means for sequentially applying a plurality of known patterns to said second processing means, thereby providing a plurality of processed known patterns each of which is uniquely associated with one of said known patterns.

Claim 12

The pattern recognition system of claim 10 wherein said unknown pattern is indicative of an unknown individual and each of said plurality of known patterns is indicative of a known individual.

Claim 13

The pattern recognition system of claim 1 further comprises: A. means for calculating the gradient of an unknown input pattern thereby providing said unknown pattern which is the gradient of said unknown input pattern; and B. means for calculating the corresponding gradient of a known input pattern thereby providing said known pattern which is the gradient of said known input pattern.

Claim 14

The pattern recognition system of claim 12 wherein: A. said first processing means further comprises: i. means for calculating a first two-dimensional gradient of said unknown pattern; ii. means for transforming said first two-dimensional gradient to a one-dimensional array; and iii. means for calculating a one-dimensional transform of said one-dimensional array thereby providing said processed unknown pattern; and B. said second processing means further comprises: i. means for calculating a second two-dimensional gradient of said known pattern; ii. means for transforming said second two-dimensional gradient to a one-dimensional array; and iii. means for calculating a one-dimensional transform of said one-dimensional array thereby providing said processed known pattern.

Claim 15

The pattern recognition system of claim 12 wherein: A. said first processing means further comprises: i. means for calculating a first two-dimensional gradient of said unknown pattern; and ii. means for calculating a two-dimensional transform of said first two-dimensional gradient thereby providing said processed unknown pattern; and B. said second processing means further comprises: i. means for calculating a second two-dimensional gradient of said known pattern; and ii. means for calculating a two-dimensional transform of said second two-dimensional gradient thereby providing said processed known pattern.

Claim 16

The pattern recognition system of claim 14 wherein said dispersive delay line includes a plurality of input transducers and a plurality of output transducers.

Claim 17

The pattern recognition system of claim 14, wherein said dispersive delay line is a two-dimensional reflective array compressor comprising a plurality of input transducers and a plurality of output transducers.

Claim 18

The pattern recognition system of claim 16 wherein said plurality of input transducers launch a surface acoustic wave to a first stack of reflectors which compresses and reduces the surface waves to a second stack of reflectors which compress and redirect the surface waves to said plurality of output transducers.

Claim 19

The pattern recognition system of claim 17, wherein said input transducers form a curved array, and said output transducers form a curved array.

Claim 20

The pattern recognition system of claim 16 wherein said plurality of input transducers launch a surface acoustic wave to a beam splitter which redirects said surface acoustic waves to a stack of reflectors which compresses and redirects said surface acoustic waves to said plurality of output arrays.

Claim 21

The pattern recognition system of claim 19 wherein said plurality of input transducers form a curved array and said plurality of output transducers form a curved array, and said reflective array compressor further comprises a collimator in front of said plurality of input transducers.

Claim 22

The pattern recognition system of claim 16 wherein said plurality of input transducers launch a surface acoustic wave to a beam splitter which redirects said surface acoustic wave to a stack of reflectors which compress and redirect said acoustic waves to an output reflector which redirects said reflected acoustic waves to said plurality of output transducers.

Claim 23

The pattern recognition system of claim 21 wherein said reflective array compressor further comprises a strip of absorbing material located between said plurality of input transducers and said plurality of output transducers to reduce undesirable coupling.

Claim 24

The pattern recognition system of claim 1: A. wherein said first processing means further comprises means for removing negative frequency values from said processed unknown pattern; and B. wherein said second processing means further comprises means for removing negative frequency values from said processed known pattern.

Claim 25

The pattern recognition system of claim 1 wherein said multiplying means further comprises means for hard-limiting said known pattern.

Claim 26

The pattern recognition system of claim 1 wherein said means for applying said product signal includes: a digital-to-analog converter responsive to said product signal which provides an analog signal indicative of said product signal; and a modulator which multiplies said analog signal with a modulation signal f m , equal to the center frequency of said dispersive delay line, thereby providing an analog product signal to said dispersive delay line.

Claim 27

The pattern recognition system of claim 25 wherein said dispersive delay line is tuned to said chirp signal f ch .

Claim 28

The pattern recognition system of claim 26 wherein said means for transforming forming further comprises a discrete Fourier transform.

Claim 29

The multiplying means of claim 1 further comprising: a first electronic memory device for storing said processed unknown pattern; a means for multiplying said chirp signal f ch by the complex conjugate of said processed known pattern, thereby providing a template; a second electronic memory device for storing said template; and a complex multiplier which multiplies output data from said first and second electronic memory devices thereby providing said product signal.

Claim 30

The pattern recognition system of claim 28, wherein said first and second electronic memory devices both comprise static random access memory (SRAM) devices.

Claim 31

The pattern recognition system of claim 28 further comprising a central processing unit which writes said processed unknown pattern to said first electronic memory and writes said template to said second electronic memory.

Claim 32

In a pattern recognition system which compares an unknown pattern with a known pattern, a method comprising the steps of: A. transforming said unknown pattern to the frequency domain to provide a transformed unknown pattern; B. transforming said known pattern to the frequency domain to provide a transformed known pattern; C. conjugating said transformed known pattern to provide a conjugated transformed known pattern; D. multiplying a chirp signal f ch by said conjugated transformed known pattern to provide a chirped conjugated transformed known pattern which is a template; E. complex multiplying said template by said transformed unknown pattern to provide a product signal; and F. applying said product signal to a dispersive delay line which provides a correlation signal whose value is indicative of the level of correlation between said known pattern and said unknown pattern.

Claim 33

In a pattern recognition system which compares an unknown pattern with a known pattern, a method comprising the steps of: A. transforming said unknown pattern to the frequency domain to provide a transformed unknown pattern; B. transforming said known pattern to the frequency domain to provide a transformed known pattern; C. conjugating said transformed known pattern to provide a conjugated transformed known pattern; D. multiplying a chirp signal f ch by said conjugated transformed known pattern to provide a chirped conjugated transformed known pattern which is a template; E. complex multiplying said template by said transformed unknown pattern to provide a product signal; and F. applying said product signal to a dispersive delay line which provides a correlation signal whose value is indicative of the level of correlation between said known pattern and said unknown pattern.

Claim 34

The method of claim 31 further comprising the steps of: sub-dividing said unknown pattern into a plurality of unknown sub-patterns; and performing steps A, E, and F on each of said unknown sub-patterns.

Claim 35

The method of claim 32, further comprising the steps of: A. calculating a mean correlation signal value and a standard deviation value, sigma, for a resulting plurality of correlation signal values; B. summing said correlation signal values between two and four sigma to create a sum value; C. calculating the square of said sum value and providing a squared signal value indicative thereof; D. dividing said squared signal value by the number of correlation signal values located between two and four sigma, and providing an estimated noise signal value indicative thereof; E. calculating, for each of said correlation signal values, a uniquely associated signal-to-noise ratio signal value by dividing said correlation signal value by said estimated noise signal value; and F. processing, for each sub-pattern, said signal-to-noise signal value and said correlation signal value to determine the level of correlation between said known and unknown patterns.

Claim 36

The method of claim 31 further comprising the steps of: providing a plurality of known patterns; and performing steps B through F on each of said known patterns.

Claim 37

The method of claim 31 further comprising the steps of: A. calculating a first gradient of an unknown input pattern to provide said unknown pattern; and B. calculating a second gradient of a known input pattern to provide said known pattern.

Claim 38

The method of claim 35 wherein: A. the step of calculating a first gradient calculates a one-dimensional gradient; and B. the step of calculating a second gradient calculates a one-dimensional gradient.

Claim 39

The method of claim 35 wherein: A. the step of calculating a first gradient calculates a two-dimensional gradient; and B. the step of calculating a second gradient calculates a two-dimensional gradient.

Claim 40

The method of claim 31 further comprising the step of removing negative frequency values from said known and unknown patterns after transforming said known and unknown patterns.

Claim 41

The method of claim 31 further comprising the step of hard limiting said template.

Claim 42

The method of claim 31 wherein the step of complex multiplying further comprises the steps of: digitally complex multiplying said template by said transformed unknown pattern to provide a digital product signal; converting said digital product signal to an analog product signal; and modulating said analog product signal with a modulation signal f m , equal to the center frequency of said dispersive delay line, to provide said product signal.

Claim 43

The method of claim 31 wherein each of said steps of transforming comprises performing a Fourier transform.

Claim 44

The method of claim 31 wherein each of said steps of transforming comprises performing a discrete Fourier transform.

Claim 45

The method of claim 31 further comprising the steps of: A. peak detecting said correlation signal to provide a peak detected signal; and B. converting said peak detected signal to a digitized peak detected signal.