Patent US5422981A successfully refreshed from Google Patents.
US5422981A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Pattern recognition method and apparatus using a neural network
Assignee: Canon Inc
•
Filing: 1991-11-18
•
Granted: 1995-06-06
•
Expires: 2011-11-18
Abstract
Abstract A method and apparatus for using a neural network to categorize patterns from pattern feature data derived from the patterns. The average of all the learning data of the neural network is subtracted from the pattern feature data, and the result is input to the input layer of the neural network. The neural network outputs a value for each category and the pattern is categorized based on these values. The neural network includes an intermediate layer whose bias is set to zero and which includes a sigmoid transfer function which is symmetric with respect to the origin.
Inventors
- 👤 Toru Niki
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-30 10:42
CPC Classifications
G; G06; G06V; G06V30/00; G06V30/10; G06V30/19; G06V30/192; G06V30/194
Claims Analysis
Total Claims: 14
Claim 1
Independent Claim
143 words
A method for training a neural network which includes at least one intermediate layer and for processing input data with the trained neural network, the method comprising the steps of: producing an average value of plural learning data; subtracting the average value from each of the plural learning data to obtain first difference data; fixing a bias for the at least one intermediate layer to zero; training the neural network, in which the bias for the at least one intermediate layer remains fixed to zero, with the first difference data; subtracting the average value from the input data to obtain second difference data; supplying the second difference data to the trained neural network, the neural network outputting a value as a function of the second difference data; and categorizing the input data on the basis of the value output from the neural network.
Claim 2
Independent Claim
143 words
A method for training a neural network which includes at least one intermediate layer and for processing input data with the trained neural network, the method comprising the steps of: producing an average value of plural learning data; subtracting the average value from each of the plural learning data to obtain first difference data; fixing a bias for the at least one intermediate layer to zero; training the neural network, in which the bias for the at least one intermediate layer remains fixed to zero, with the first difference data; subtracting the average value from the input data to obtain second difference data; supplying the second difference data to the trained neural network, the neural network outputting a value as a function of the second difference data; and categorizing the input data on the basis of the value output from the neural network.
Claim 3
Dependent Claim
37 words
A method according to claim 1, further comprising the step of applying a sigmoid transfer function which is symmetric with respect to the origin to data to produce the output of the at least one intermediate layer.
Claim 4
Dependent Claim
41 words
A method according to claim 1, wherein said step of subtracting the average value from the input data and said step of supplying the second difference data are both performed each time input data is processed with the trained neural network.
Claim 5
Independent Claim
149 words
A method for categorizing input data into one of plural categories using a neural network which includes at least one intermediate layer, said method comprising the steps of: producing an average value of plural learning data; subtracting the average value from each of the plural learning data to obtain first difference data; fixing a bias for the at least one intermediate layer to zero; training the neural network, in which the bias for the at least one intermediate layer remains fixed to zero, with the first difference data; subtracting the average value from the input data to obtain second difference data; supplying the second difference data to the trained neural network; processing the second difference data with the trained neural network; receiving a value corresponding to each category from the trained neural network; and determining one of the plural categories for the input data based on the received values.
Claim 6
Independent Claim
149 words
A method for categorizing input data into one of plural categories using a neural network which includes at least one intermediate layer, said method comprising the steps of: producing an average value of plural learning data; subtracting the average value from each of the plural learning data to obtain first difference data; fixing a bias for the at least one intermediate layer to zero; training the neural network, in which the bias for the at least one intermediate layer remains fixed to zero, with the first difference data; subtracting the average value from the input data to obtain second difference data; supplying the second difference data to the trained neural network; processing the second difference data with the trained neural network; receiving a value corresponding to each category from the trained neural network; and determining one of the plural categories for the input data based on the received values.
Claim 7
Dependent Claim
36 words
A method according to claim 4, further comprising the step applying a sigmoid transfer function which is symmetric with respect to the origin to data to produce the output of the at least one intermediate layer.
Claim 8
Dependent Claim
157 words
A pattern recognition apparatus for categorizing an input pattern into one of plural categories, comprising: feature extracting means for extracting feature data from the input pattern; storing means for storing an average value of all the feature data extracted by said feature extracting means from a plurality of learning patterns; subtracting means for subtracting the average value stored in said storing means from the feature data extracted by the feature extracting means to obtain difference data; a neural network which includes at least one intermediate layer in which a bias is fixed to zero, said neural network being trained with data obtained by subtracting the average value from each of the feature data for the plural learning patterns, said neural network receiving the difference data from said subtracting means and outputting a value for each of the plural categories; and categorizing means for categorizing the input pattern on the basis of the outputs from said neural network.
Claim 9
Dependent Claim
157 words
A pattern recognition apparatus for categorizing an input pattern into one of plural categories, comprising: feature extracting means for extracting feature data from the input pattern; storing means for storing an average value of all the feature data extracted by said feature extracting means from a plurality of learning patterns; subtracting means for subtracting the average value stored in said storing means from the feature data extracted by the feature extracting means to obtain difference data; a neural network which includes at least one intermediate layer in which a bias is fixed to zero, said neural network being trained with data obtained by subtracting the average value from each of the feature data for the plural learning patterns, said neural network receiving the difference data from said subtracting means and outputting a value for each of the plural categories; and categorizing means for categorizing the input pattern on the basis of the outputs from said neural network.
Claim 10
Dependent Claim
23 words
An apparatus according to claim 6, wherein said neural network includes a sigmoid transfer function which is symmetric with respect to the origin.
Claim 11
Dependent Claim
117 words
A neural network for categorizing input data into one of plural categories, comprising: an input layer for inputting a value obtained by subtracting from the input data an average value of plural learning data; at least one intermediate layer for processing data from the input layer, said at least one intermediate layer having a bias fixed to zero and a sigmoid transfer function which is symmetric with respect to the origin; and an output layer for receiving data from said at least one intermediate layer and outputting a value according to each category, wherein said neural network is trained with data obtained by subtracting the average value of the plural learning data from the plural learning data.
Claim 12
Dependent Claim
117 words
A neural network for categorizing input data into one of plural categories, comprising: an input layer for inputting a value obtained by subtracting from the input data an average value of plural learning data; at least one intermediate layer for processing data from the input layer, said at least one intermediate layer having a bias fixed to zero and a sigmoid transfer function which is symmetric with respect to the origin; and an output layer for receiving data from said at least one intermediate layer and outputting a value according to each category, wherein said neural network is trained with data obtained by subtracting the average value of the plural learning data from the plural learning data.
Claim 13
Dependent Claim
184 words
A pattern recognition apparatus for categorizing a visually perceptible pattern into one of plural categories, comprising: converting means for converting the visually perceptible pattern into an electrical signal; preprocessing means for normalizing and binarizing the electrical signal; feature extracting means for extracting feature data from the normalized and binarized electrical signal; storing means for storing an average value of all the feature data extracted from normalized and binarized electrical signals for plural learning patterns; subtracting means for subtracting the average value stored in said storing means from the feature data extracted by said feature extracting means to form difference data; a neural network which includes at least one intermediate layer in which a bias is fixed to zero, said neural network being trained with the data obtained by subtracting the average value from each of the feature data for the plural learning patterns, said neural network receiving the difference data from said subtracting means and outputting a value for each of the plural categories; and outputting means for outputting one of the plural categories on the basis of the outputs from said neural network.
Claim 14
Dependent Claim
184 words
A pattern recognition apparatus for categorizing a visually perceptible pattern into one of plural categories, comprising: converting means for converting the visually perceptible pattern into an electrical signal; preprocessing means for normalizing and binarizing the electrical signal; feature extracting means for extracting feature data from the normalized and binarized electrical signal; storing means for storing an average value of all the feature data extracted from normalized and binarized electrical signals for plural learning patterns; subtracting means for subtracting the average value stored in said storing means from the feature data extracted by said feature extracting means to form difference data; a neural network which includes at least one intermediate layer in which a bias is fixed to zero, said neural network being trained with the data obtained by subtracting the average value from each of the feature data for the plural learning patterns, said neural network receiving the difference data from said subtracting means and outputting a value for each of the plural categories; and outputting means for outputting one of the plural categories on the basis of the outputs from said neural network.
References Cited (Backward Citations)
US4941122A
Cited
US5119438A
Cited
US5060278A
Cited
US5058179A
Cited
US5091780A
Cited
US5142612A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1998-06-05 | 1998-12-05 | 1999-06-06 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2002-06-05 | 2002-12-05 | 2003-06-06 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2006-06-05 | 2006-12-05 | 2007-06-06 | $7700.00 | Window Closed / Lapsed |