The pattern recognition or recognition of reasons is a under-field of the machine Learning and can be defined like the action to take in entered of the raw data in order to make a decision based on the category of these data .

The goal of the pattern recognition is to classify data by basing on knowledge a priori or information Statistique drawn from the reasons. The reasons to be classified are usually sets of measures or observations which define points in a multidimensional space suitable.

Methods

The recognition of reasons can be carried out by:

• a Network of neurons
• a Statistical analysis
• the use of model of Markov hidden
• a search for isomorphism of graphs or subgraphs

The required forms are often geometrical, describable forms by a mathematical formula, such as:

• Circle or curved ellipse
• of Bézier, splines
• right

It can also be of more complex nature:

• letter
• Figure
• Digital fingerprint

The algorithms of recognition can work on images in black and white, with in white contours of the objects being in the image. These images are the fruit of algorithms of Détection of contour S. They can also work on zones of the image preset resulting from the segmentation of the image.

Methods of pattern recognition:

• Put in correspondence of graphs
• Method Linear Bayesienne
• Parametric Estimate
• Classifying
• Local Network of neurons
• x-ray feature
• SVM: Support Vector Machine
• Polytôpes of constraint
• Method of the hypercubes

A well-known algorithm for the detection of forms, the Transformed of Hough, is a method of estimate parametric.

Global method

This method characterizes a form and extract of the parameters characteristic of the object and compare them by a method of classification or mapping with a base of training. By this method, it is impossible to extract several forms from the same image without preprocessing.

Multiple method starting from point of interest

In this approach, one extracts from the points characteristic of objects like the corners via the detectors of Harris then one extracts from the characteristics in the vicinity of this point. With these characteristics, it is possible to extract several objects and to make the recognition of those via a classifior.

Applications

• Play of go per computer
• Voice recognition
• Analysis and Recognition of images
• Classification of documents
• Recognition of the manuscript writing

Refer

Random links:

Day memorial | Chaumont-in front of-Damvillers | Pignans | Gerard of the cinema | Trust (undertaken)