Jump to content

User:Thepigdog/Perception recognition and object model

From Wikipedia, the free encyclopedia

Perception is the first stage of data processing. The brain interprets the world as a collection of objects. Perception is then interpreting input data using an object model.

The second stage of perception is recognition. Recognition is classifying objects by how similar objects are to each other. This similarity map is then interpreted a tree structure. Objects may fit into multiple inheritance trees. This is similar to the multiple inheritance class structure in Object oriented programming.

Objects are classified by nouns and adjectives. The role of a noun and adjective are similar.

Once objects have been identified, and specific points on an object may be identified then movements and actions may be perceived, through applying recognition to motion capture.

Actions are sequences of motion usually starting from rest and ending in rest. Actions are classified by verbs and adverbs.

Object model

[edit]

The object model classified "objects" and "actions".

  • An object has a contiguous surface, and may be perceived by sight, sound, and touch.
  • An action is a movement or sequence of movements made by an object, or parts of an object.

The purpose of this model is not to classify natural language. Instead the object model reflects certain aspects of the world that are represented in natural language. Natural language is more complex than it needs to be, in order to utilize sophisticated pattern matching and recognition abilities, but poor abilities to process recursive minimalist structures (and match brackets). So natural language is not designed with a symmetrical minimalist structure. But internally the concepts are understood logically, even if the language structures are haphazard.

Objects

[edit]

A dog called named Fido may be classified by,

Attribute Color Species Age
Class Dog
Class Black Terrier Young
Identity Fido

Fido is classified by,

  • Attribute
  • Class
  • Identity
Object class Grammar Phrase
Identity Proper noun Fido
Membership of primary attribute determiner noun A dog
Membership of primary and secondary attributes determiner adjective noun A red dog

species(fido) = dog and breed(fido) = terrier and color(fido) = red and age(fido) = young


Components

[edit]

Actions

[edit]

An action called "walk" may be classified by,

Attribute Transport Speed Directionality
Class walk fast random
Identity Fido's standard walk

Fido's walk is classified by,

  • Attribute
  • Class
  • Identity
Action class Grammar Phrase
Membership of action attribute verb walk
Membership of primary and secondary action attributes adverb verb random walk

Time model

[edit]

Objects exist in time. Objects change over time. So any statement about an object must be qualified by time.

As a default, a statement made about an object remains unchanged, unless some action changes it. This makes the imperative model the most natural method of representing objects. State holder types, may be used to link the imperative model to the mathematical model.

Perception

[edit]

Perception is the first stage of processing of data. The role of perception is to identify objects and attach information to them.

Vision systems

[edit]

Stereo vision systems match points in one image with points in a second image, by minimizing an information measure, so as to find the best fit of one image to another, and derive a disparity map. The disparity map maybe used to determine the distances from the observer of pixels in the image. The result encodes the distance from the observe position of pixels in the image.

Methods of implementation

[edit]

The minimization problem is NP-complete. This means a general solution to this problem will take an unthinkably long time to reach a solution. However methods exist for computers based on heuristics that approximate the result in a reasonable amount of time. Also methods exist based on neural networks [1]. Efficient implementation of stereoscopic vision is an area of active research.

Sound systems

[edit]

Recognition

[edit]

Image/object recognition

[edit]

Object recognition is the means of identifying and classifying objects. It forms the basis for identifying physical nouns and adjectives.

Motion capture recognition

[edit]

Motion capture recognition is the means of of identify the physical verbs.

Sound recognition

[edit]
  1. ^ WANG, JUNG-HUA; HSIAO, CHIH-PING (1999). Proc. Natl. Sci. Counc. ROC(A). 23 (5): 665-678. {{cite journal}}: Missing or empty |title= (help)