Method and apparatus for processing information and a method and apparatus for executing a work instruction

• US 5,845,050 A
• Filed: 01/11/1995
• Issued: 12/01/1998
• Est. Priority Date: 02/28/1994
• Status: Expired due to Term

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Sekiguchi, Minoru
• Primary Examiner(s): Downs, Robert W.
• Assistant Examiner(s): Smith, Jeffrey S.
• Application Number:
  US08/371,430
• Time in Patent Office: 1,420 Days
• Field of Search: 395/22, 395/23, 395/24, 395/10, 395/11, 395/27, 395/94, 382/153
• US Class Current: 706/20
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G06N 3/042 Knowledge-based neural networks; Logical representations of neural networks

Abstract

A first information transforming unit, capable of learning, transforms state information acquired from an object into a first intermediate representation. A second information transforming unit, capable of learning, transforms knowledge information derived, from attributes of the object, by a human being into a second intermediate representation. A fusing unit produces a common intermediate representation using the first intermediate representation and the second intermediate representation. The common intermediate representation is used as a teacher signal in order to train the first and second information transforming units. Thereafter, an intermediate representation made by transforming state information acquired from a certain object will be consistent with an intermediate representation made by transforming knowledge information derived from the object by a human being.

Claims

1. A learning method for a pattern recognition apparatus, comprising:

acquiring state information of an object;

providing a first neural network to learn to transform the state information into a first intermediate representation;

providing a second neural network to learn to transform input knowledge of the object into a second intermediate representation;

fusing said first intermediate representation and said second intermediate representation to produce a common intermediate representation of the object; and

applying said common intermediate representation as a teacher signal to said first and second neural networks to train said first and second neural networks.

2. An information processing method according to claim 1, wherein, said fusing produces said common intermediate representation, by an OR operation on said two intermediate representations.

3. A pattern recognition method, comprising:

acquiring state information of an object;

providing a first neural network to transform the state information into a first intermediate representation;

providing a second neural network to transform input knowledge information of the object into a second intermediate representation;

fusing said first intermediate representation and said second intermediate representation to produce a common intermediate representation of the object, and

training said first and second neural networks by applying a teacher signal derived from said common intermediate representation;

providing said trained first neural network to transform the state information of the object in to a third intermediate representation;

providing said trained second neural network to transform the input knowledge information to be compared to said state information of the object into a fourth intermediate representation; and

comparing said third intermediate representation with said fourth intermediate representation to determine whether a concept represented by said state information of the object and a concept represented by said knowledge information to be compared with said state information of said object are consistent with each other.

4. A method of executing a work instruction, comprising steps of:

teaching a neural network to recognize an object, including the steps of;

acquiring state information of the object via a detection unit;

providing a first neural network to transform the state information of the object into a first intermediate representation,

providing a second neural network to transform input knowledge information of the object into a second intermediate representation,

fusing said first intermediate representation and said second intermediate representation to produce a common intermediate representation of the object, and

training said first and second neural networks by applying a teacher signal derived from said common intermediate representation;

inputting a work instruction;

interpreting said input work instruction to identify separate units of information;

retrieving knowledge information concerning a concept represented by at least one of the separate units of information;

allowing said trained second neural network to transform said knowledge information into a fourth intermediate representation;

allowing said trained first neural network to transform said state information of an object into a third intermediate representation; and

comparing said third intermediate representation with said fourth intermediate representation to determine whether a concept represented by said state information of the object and a concept represented by said knowledge information are consistent with each other.

5. A pattern recognition apparatus, comprising:

a first neural network, for transforming input state information into a first intermediate representation;

a second neural network for transforming input knowledge information into a second intermediate representation;

a fusing means for fusing said first intermediate representation and said second intermediate representation, applying said common intermediate representation as a teacher signal to said first and second neural networks, and thus training said first and second neural networks; and

a comparing means for comparing a third intermediate representation which is transformed from state information of an object by said trained first neural network with a fourth intermediate representation which is transformed from knowledge information of an object to be detected by said trained second neural network, and determining whether or not the object and the object to be determined are consistent with each other.

6. A pattern recognition apparatus according to claim 5, further comprising:

means for interpreting an input work instruction and splitting it in units of information; and

a means for retrieving knowledge information concerning a concept represented by said split information.