Patent US6038338A successfully refreshed from Google Patents.
US6038338A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Hybrid neural network for pattern recognition
Assignee: United States Department of the Navy
•
Filing: 1997-02-03
•
Granted: 2000-03-14
•
Expires: 2017-02-03
Abstract
Abstract A system and a method for recognizing patterns comprises a first stage forxtracting features from inputted patterns and for providing topological representations of the characteristics of the inputted patterns and a second stage for classifying and recognizing the inputted patterns. The first stage comprises two one-layer neural networks and the second stage comprises a feedforward two-layer neural network. Supplying signals representative of a set of inputted patterns to the input layers of the first and second neural networks, training the first and second neural networks using a competitive learning algorithm, and generating topological representations of the input patterns using the first and second neural networks The method further comprises providing a third neural network for classifying and recognizing the inputted patterns and training the third neural network with a back-propagation algorithm so that the third neural network recognizes at least one interested pattern.
Inventors
- 👤 Chung T. Nguyen
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 14:38
CPC Classifications
G; G06; G06F; G06F18/00; G06F18/20; G06F18/21; G06F18/213; G06F18/2137; G06F18/24
Claims Analysis
Total Claims: 17
Claim 1
Independent Claim
145 words
A system for recognizing patterns which comprises: a first stage for extracting features from inputting patterns and for providing topological representations of the characteristics of said inputted patterns; said first stage comprising a first neural network for receiving a first set of input information and for generating a first set of topographical representations from said first set of input information and a second neural network for receiving a second set of input information, which second set of input information is different from said first set of input information, and for generating a second set of topographical representations from said second set of input information; a second stage for classifying and recognizing said inputted patterns; and said second stage comprising a feedforward two-layer neural network which uses said topological representations of said input patterns generated by said first and second neural networks for training said system.
Claim 2
Independent Claim
145 words
A system for recognizing patterns which comprises: a first stage for extracting features from inputting patterns and for providing topological representations of the characteristics of said inputted patterns; said first stage comprising a first neural network for receiving a first set of input information and for generating a first set of topographical representations from said first set of input information and a second neural network for receiving a second set of input information, which second set of input information is different from said first set of input information, and for generating a second set of topographical representations from said second set of input information; a second stage for classifying and recognizing said inputted patterns; and said second stage comprising a feedforward two-layer neural network which uses said topological representations of said input patterns generated by said first and second neural networks for training said system.
Claim 3
Dependent Claim
41 words
A system according to claim 1 wherein each of said first and second neural networks comprises a one-layer network having an input layer formed by a plurality of input neurons and an output layer formed by a plurality of output neurons.
Claim 4
Dependent Claim
39 words
A system according to claim 2 wherein the sets of input information contain information about a plurality of properties in said input signal and each of said sets of input information contain information about different ones of said properties.
Claim 5
Dependent Claim
30 words
A system according to claim 3 wherein each input neuron in said input layer computes information which it receives and presents the results to each neuron in said output layer.
Claim 6
Independent Claim
205 words
A system for recognizing patterns which comprises: first means for extracting features from inputted patterns and for providing topological representations of the characteristics of said inputted patterns; said first means comprising a first neural network for receiving a first set of input information and for generating a first set of topographical representations from said first set of input information and a second neural network for receiving a second set of input information different from said first set of input information and for generating a second set of topographical representations from said second set of input information; second means for classifying and recognizing inputted patterns; said second means comprising a feedforward two-layer neural network which uses said topological representations of said input patterns generated by said first and second neural networks for training said system; each of said first and second neural networks comprises a one-layer network having an input layer formed by a plurality of input neurons and an output layer formed by a plurality of output neurons; and said input neurons are connected to said output neurons by feedforward connections and wherein said output neurons are laterally connected so that each output neuron tends to inhibit each neuron to which it is laterally connected.
Claim 7
Independent Claim
205 words
A system for recognizing patterns which comprises: first means for extracting features from inputted patterns and for providing topological representations of the characteristics of said inputted patterns; said first means comprising a first neural network for receiving a first set of input information and for generating a first set of topographical representations from said first set of input information and a second neural network for receiving a second set of input information different from said first set of input information and for generating a second set of topographical representations from said second set of input information; second means for classifying and recognizing inputted patterns; said second means comprising a feedforward two-layer neural network which uses said topological representations of said input patterns generated by said first and second neural networks for training said system; each of said first and second neural networks comprises a one-layer network having an input layer formed by a plurality of input neurons and an output layer formed by a plurality of output neurons; and said input neurons are connected to said output neurons by feedforward connections and wherein said output neurons are laterally connected so that each output neuron tends to inhibit each neuron to which it is laterally connected.
Claim 8
Dependent Claim
55 words
A system according to claim 1 wherein said two-layer neural network comprises an input layer formed by a plurality of input neurons, a hidden layer formed by a plurality of neurons which are fewer in number than the number of neurons forming said input layer, and an output layer formed by a single output neuron.
Claim 9
Dependent Claim
28 words
A system according to claim 6 wherein said output neuron generates a first output signal representative of a match or a second output signal representative of a non-match.
Claim 10
Independent Claim
147 words
A method for recognizing patterns which comprises: providing first and second neural networks each having an input layer formed by a plurality of input neurons and an output layer formed by a plurality of output neurons; supplying signals representative of features of a set of inputted patterns to the input layers of said first and second neural networks with a first set of inputted patterns being supplied to the input layer of the first neural network and a second set of inputted patterns different from the first set being supplied to the input layer of the second neural network; training said first and second neural networks using a competitive learning algorithm; generating topological representations of the inputted patterns using said first and second neural networks; providing a third neural network for classifying and recognizing said inputted patterns; and inputting said topological representations into said third neural network.
Claim 11
Independent Claim
147 words
A method for recognizing patterns which comprises: providing first and second neural networks each having an input layer formed by a plurality of input neurons and an output layer formed by a plurality of output neurons; supplying signals representative of features of a set of inputted patterns to the input layers of said first and second neural networks with a first set of inputted patterns being supplied to the input layer of the first neural network and a second set of inputted patterns different from the first set being supplied to the input layer of the second neural network; training said first and second neural networks using a competitive learning algorithm; generating topological representations of the inputted patterns using said first and second neural networks; providing a third neural network for classifying and recognizing said inputted patterns; and inputting said topological representations into said third neural network.
Claim 12
Dependent Claim
34 words
A method according to claim 8 further comprising: training said third neural network with a back-propagation algorithm and said inputted topological representations so that said third neural network recognizes at least one interested pattern.
Claim 13
Dependent Claim
47 words
A method according to claim 9 wherein said third neural network training step comprises using an algorithm based on the error correction learning rule: ##EQU4## where η is a constant that determines the rate of learning, Δw ji is the correction weight, and ε is the error.
Claim 14
Dependent Claim
32 words
A method according to claim 8 wherein said training step comprises training said output neurons to compete among themselves to be the one active output neuron at a particular instant of time.
Claim 15
Dependent Claim
147 words
A hybrid neural network system for pattern recognition which comprises: a first feature extraction stage for extracting features from inputting patterns and for providing topological representations of the characteristics of said inputted patterns; said first feature attraction stage comprising a first neural network for receiving a first set of input information and for generating a first set of topographical representations from said first set of input information and a second neural network for receiving a second set of input information different from said first set of input information and for generating a second set of topographical representations from said second set of input information; a second classification stage for classifying and recognizing said inputted patterns; and said second classification stage comprising a feedforward two-layer neural network which uses said topological representations of said input patterns generated by said first and second neural networks for training said system.
Claim 16
Dependent Claim
147 words
A hybrid neural network system for pattern recognition which comprises: a first feature extraction stage for extracting features from inputting patterns and for providing topological representations of the characteristics of said inputted patterns; said first feature attraction stage comprising a first neural network for receiving a first set of input information and for generating a first set of topographical representations from said first set of input information and a second neural network for receiving a second set of input information different from said first set of input information and for generating a second set of topographical representations from said second set of input information; a second classification stage for classifying and recognizing said inputted patterns; and said second classification stage comprising a feedforward two-layer neural network which uses said topological representations of said input patterns generated by said first and second neural networks for training said system.
Claim 17
Dependent Claim
46 words
The hybrid neural network of claim 12, further comprising: said first and second neural networks each being of the type to be trained using an unsupervised competitive learning algorithm; and said feedforward two-layer neural network being of the type to be trained using a back-propagation algorithm.
References Cited (Backward Citations)
US5023833A
Cited
US5426745A
Cited
US5402519A
Cited
US5105468A
Cited
US5390284A
Cited
US5293456A
Cited
US5712922A
Cited
US5581660A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2003-03-14 | 2003-09-13 | 2004-03-14 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2007-03-14 | 2007-09-13 | 2008-03-14 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2011-03-14 | 2011-09-13 | 2012-03-14 | $7700.00 | Window Closed / Lapsed |