Artificial intelligence systems for classifying events, objects and situations

• US 7,233,936 B1
• Filed: 06/28/2000
• Issued: 06/19/2007
• Est. Priority Date: 07/01/1999
• Status: Expired due to Term

• Current Assignee: Commissariat a L'Energie Atomique (French Alternative Energies and Atomic Energy Commission)
• Original Assignee: Commissariat a L'Energie Atomique (French Alternative Energies and Atomic Energy Commission)
• Inventors: Carceles, Ste, Muller, Denis
• Primary Examiner(s): DAVIS, GEORGE B
• Application Number:
  US10/009,335
• Time in Patent Office: 2,547 Days
• Field of Search: 706/2, 706/52
• US Class Current: 706/52
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06N 5/045 Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
• CPC Class Codes: G06N 5/048 Fuzzy inferencing

Abstract

The present invention relates to a system of artificial intelligence for the classification of events, objects or situations from signals and discriminant parameters produced by models comprising at least one processing branch comprising a fuzzy expert system making a decision according to high level properties and discriminant parameters of low level extracted from signals by signal processing type procedures, and capable of explaining its decision to the user through the intermediary of rules selected by order of applicability.

Claims

1. An artificial intelligence system for classification of events of geophysical recordings, comprising:

a neuro-fuzzy classifier branch for making a decision from high level properties of the events of geophysical recordings and lower level parameters extracted from signals by signal processing type procedures;

a fuzzy expert system branch for generating a decision and explaining the decision to a user through an intermediary of rules selected by order of applicability of the events of geophysical recordings; and

a neural network branch with local connections and shared weights, formed by banks of non-linear adaptable filters, the neural network is extracting relevant information for time-frequency representations from signals corresponding to the events of geophysical recordings,

wherein the branches are automatically configured by statistical learning of the events of geophysical recordings in a database.

2. The system according to claim 1 wherein the fuzzy expert system with a gradient decrease is carried out on the parameters:

x=y/σ

s=ln(2σ

²)

r=ln(ρ

)

d

with;

y;

position of fuzzy sets of premises

σ

;

width of fuzzy sets of premises

ρ

;

weights of rules

d;

degree of activation of each class for each rule

x;

convergence value of a position of the fuzzy sets of premises

s;

convergence modification of the width of fuzzy sets of premises

r;

constant weights of the rules.

3. The system according to claim 1, in which the high level properties are a localisation, a magnitude, a time and a date.