Patent US5859925A successfully refreshed from Google Patents.
US5859925A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Classifying system having a single neural network architecture for multiple input representations
Assignee: Apple Inc
•
Filing: 1995-08-08
•
Granted: 1999-01-12
•
Expires: 2015-08-08
Abstract
Abstract A classification system is provided for combining multiple input representations by a single neural network architecture. In such a classification system having a single neural network architecture, classification channels corresponding to various input representations may be integrated through their own and shared hidden layers of the network to produce highly accurate classification. The classification system is particularly applicable to character classifying applications which use stroke and character image features as the main classification criteria, along with scalar features such as stroke count and aspect ratio features as secondary classification. The classification channels corresponding to the scalar features may be cross wired to the classification channels corresponding to the main input representations for further improving the accuracy of the classification output. Because a single neural network architecture is used, only one, standard training technique is needed for this classification system, special data handling is minimized, and the training time can be reduced, while highly accurate classification is achieved.
Inventors
- 👤 Larry S. Yaeger
- 👤 Brandyn Webb
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-29 17:21
CPC Classifications
G; G01; G01V; G01V1/00; G01V1/28; G01V1/30; G06; G06F; G06F18/00; G06F18/20; G06F18/25; G06F18/253; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/0499; G06N3/08; G06N3/09
Claims Analysis
Total Claims: 40
Claim 1
Independent Claim
152 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of non-scalar representations of a character to be classified, and at least one scalar representation defining a character recognition feature which is capable of distinguishing different characters from one another; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a respective one of said non-scalar representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives said scalar representation and produces output values that are coupled to said hidden layers of said primary classification channels.
Claim 2
Dependent Claim
152 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of non-scalar representations of a character to be classified, and at least one scalar representation defining a character recognition feature which is capable of distinguishing different characters from one another; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a respective one of said non-scalar representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives said scalar representation and produces output values that are coupled to said hidden layers of said primary classification channels.
Claim 3
Dependent Claim
35 words
The classification system of claim 1 wherein the output values from said secondary channel are also directly coupled to said output layer of said neural network in addition to being coupled to said hidden layers.
Claim 4
Dependent Claim
37 words
The classification system of claim 1 wherein said input device develops a plurality of scalar representations of an input character, and said neural network includes a plurality of secondary classification channels respectively associated with said scalar representations.
Claim 5
Dependent Claim
16 words
The classification system of claim 1, wherein one of said non-scalar representations comprises character image features.
Claim 6
Dependent Claim
16 words
The classification system of claim 4 wherein another one of said non-scalar representations comprises stroke features.
Claim 7
Dependent Claim
12 words
The classification system of claim 5 wherein said stroke features include slope.
Claim 8
Dependent Claim
12 words
The classification system of claim 5 wherein said stroke features include curvature.
Claim 9
Dependent Claim
15 words
The classification system of claim 1 wherein one of said non-scalar representations comprises stroke features.
Claim 10
Dependent Claim
13 words
The classification system of claim 1 wherein said scalar representation comprises stroke count.
Claim 11
Dependent Claim
13 words
The classification system of claim 1 wherein said scalar representation comprises aspect ratio.
Claim 12
Independent Claim
128 words
A method for classifying input characters, comprising the steps of: developing a plurality of non-scalar representations of a character to be classified, and at least one scalar representation of said character defining a character recognition feature which is capable of distinguishing different characters from one another; separately processing said non-scalar representations in respective classification channels of a neural network, each of which has at least one hidden layer of processing nodes, to produce intermediate values indicative of an input character; coupling values related to said scalar representation to the hidden layer of said classification channels; and processing the intermediate values from said classification channels in an output layer of processing nodes of said neural network, to produce output values representative of the probable classification of the input character.
Claim 13
Independent Claim
128 words
A method for classifying input characters, comprising the steps of: developing a plurality of non-scalar representations of a character to be classified, and at least one scalar representation of said character defining a character recognition feature which is capable of distinguishing different characters from one another; separately processing said non-scalar representations in respective classification channels of a neural network, each of which has at least one hidden layer of processing nodes, to produce intermediate values indicative of an input character; coupling values related to said scalar representation to the hidden layer of said classification channels; and processing the intermediate values from said classification channels in an output layer of processing nodes of said neural network, to produce output values representative of the probable classification of the input character.
Claim 14
Dependent Claim
34 words
The method of claim 11 wherein the values related to said scalar representation are also directly coupled to said output layer of said neural network in addition to being coupled to said hidden layer.
Claim 15
Dependent Claim
34 words
The method of claim 11 further including the steps of developing a plurality of scalar representations of a character, and coupling values related to each of said plural scalar representations to said hidden layers.
Claim 16
Dependent Claim
35 words
The method of claim 11 further including the steps of back propagating an error value, derived from said probable classification, through said output layer and each of said classification channels to train said neural network.
Claim 17
Dependent Claim
16 words
The classification system of claim 11, wherein one of said non-scalar representations comprises character image features.
Claim 18
Dependent Claim
16 words
The classification system of claim 15 wherein another one of said non-scalar representations comprises stroke features.
Claim 19
Dependent Claim
12 words
The classification system of claim 16 wherein said stroke features include slope.
Claim 20
Dependent Claim
12 words
The classification system of claim 16 wherein said stroke features include curvature.
Claim 21
Dependent Claim
15 words
The classification system of claim 11 wherein one of said non-scalar representations comprises stroke features.
Claim 22
Dependent Claim
13 words
The classification system of claim 11 wherein said scalar representation comprises stroke count.
Claim 23
Dependent Claim
13 words
The classification system of claim 11 wherein said scalar representation comprises aspect ratio.
Claim 24
Dependent Claim
154 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of representations of a character to be classified, wherein each of said representations defines a character recognition feature which is capable of distinguishing different characters from one another; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a different one of said representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives another one of said representations of said character and produces output values that are coupled to said hidden layers of said primary classification channels.
Claim 25
Dependent Claim
154 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of representations of a character to be classified, wherein each of said representations defines a character recognition feature which is capable of distinguishing different characters from one another; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a different one of said representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives another one of said representations of said character and produces output values that are coupled to said hidden layers of said primary classification channels.
Claim 26
Dependent Claim
35 words
The classification system of claim 22 wherein the output values from said secondary channel are also directly coupled to said output layer of said neural network in addition to being coupled to said hidden layers.
Claim 27
Dependent Claim
30 words
The classification system of claim 22 wherein said neural network includes a plurality of secondary classification channels associated with respective ones of said representations and coupled to said hidden layers.
Claim 28
Dependent Claim
15 words
The classification system of claim 22, wherein one of said representations comprises character image features.
Claim 29
Dependent Claim
15 words
The classification system of claim 25 wherein another one of said representations comprises stroke features.
Claim 30
Dependent Claim
12 words
The classification system of claim 26 wherein said stroke features include slope.
Claim 31
Dependent Claim
12 words
The classification system of claim 26 wherein said stroke features include curvature.
Claim 32
Dependent Claim
14 words
The classification system of claim 22 wherein one of said representations comprises stroke features.
Claim 33
Dependent Claim
14 words
The classification system of claim 22 wherein one of said representations comprises stroke count.
Claim 34
Dependent Claim
13 words
The classification system of claim 22 wherein one of said representations comprises ratio.
Claim 35
Dependent Claim
151 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of representations of a character to be classified, including at least one scalar representation of said character; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a respective one of said representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives said scalar representation and produces output values that are coupled to said hidden layers of said primary classification channels and are also directly coupled to said output layer of said neural network.
Claim 36
Dependent Claim
151 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of representations of a character to be classified, including at least one scalar representation of said character; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a respective one of said representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives said scalar representation and produces output values that are coupled to said hidden layers of said primary classification channels and are also directly coupled to said output layer of said neural network.
Claim 37
Independent Claim
139 words
A method for classifying input characters, comprising the steps of: developing a plurality of representations of a character to be classified, including at least one scalar representation of said character; separately processing said representations in respective classification channels of a neural network, each of which has at least one hidden layer of processing nodes, to produce intermediate values indicative of an input character; coupling values related to said scalar representation to the hidden layer of each of said classification channels; processing the intermediate values from said classification channels in an output layer of processing nodes of said neural network, to produce output values representative of the probable classification of the input character; and directly coupling the values related to said scalar representation to said output layer of said neural network in addition to coupling them to said hidden layer.
Claim 38
Independent Claim
139 words
A method for classifying input characters, comprising the steps of: developing a plurality of representations of a character to be classified, including at least one scalar representation of said character; separately processing said representations in respective classification channels of a neural network, each of which has at least one hidden layer of processing nodes, to produce intermediate values indicative of an input character; coupling values related to said scalar representation to the hidden layer of each of said classification channels; processing the intermediate values from said classification channels in an output layer of processing nodes of said neural network, to produce output values representative of the probable classification of the input character; and directly coupling the values related to said scalar representation to said output layer of said neural network in addition to coupling them to said hidden layer.
Claim 39
Dependent Claim
147 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of representations of a character to be classified; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a different one of said representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives another one of said representations of said character and produces output values that are coupled to said hidden layers of said primary classification channels and are also directly coupled to said output layer of said neural network.
Claim 40
Dependent Claim
147 words
A classification system for classifying input characters, comprising: an input device which develops a plurality of representations of a character to be classified; and a neural network having: a plurality of primary classification channels each having at least one hidden layer of processing nodes, each of said primary classification channels processing a different one of said representations to produce intermediate values indicative of an input character; an output layer of processing nodes which receives the intermediate values from said plurality of primary classification channels to produce output values representative of the probable classification of an input character; and a secondary classification channel having at least one input node which receives another one of said representations of said character and produces output values that are coupled to said hidden layers of said primary classification channels and are also directly coupled to said output layer of said neural network.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2002-01-11 | 2002-07-13 | 2003-01-12 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2006-01-11 | 2006-07-13 | 2007-01-12 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2010-01-11 | 2010-07-13 | 2011-01-12 | $7700.00 | Window Closed / Lapsed |