Pattern recognition

Document Sample
Pattern recognition Powered By Docstoc
					                Pattern Recognition

Prepared By :
What is pattern recognition?
Pattern recognition system
Process diagram of PR system
Description and Classification task
Feature vector
Approaches to Pattern recognition
 Human Perception
Humans have developed highly sophisticated skills for
 sensing their environment and taking actions according
 to what they observe, e.g.,
 recognizing a face
 understanding spoken words
 reading handwriting
 distinguishing fresh food from its smell
->Similar capabilities to machines.
What is Pattern Recognition?
 Pattern recognition is a sub-topic of machine
  learning. PR is the science that concerns the
  description or classification (recognition) of
  measurements. It can be defined as:
   "the act of taking in raw data and taking an
     action based on the category of the data".
   “the assignment of a physical object or event to
     one of several prespecified categories”.
 A pattern is an object, process or event that can
  be given a name.
 A pattern class (or category) is a set of patterns
  sharing common attributes and usually
  originating from the same source.
 During recognition (or classification) given
  objects are assigned to prescribed classes.
 A classifier is a machine which performs
 Pattern recognition aims to classify data
  (patterns) based on either a priori knowledge or
  on statistical information extracted from the
 The patterns to be classified are usually groups
  of measurements or observations, defining
  points in an appropriate multidimensional space.
 PR techniques are an important component of
  intelligent systems and are used for :-
   Decision making
   Object & pattern classification
   Data preprocessing
Pattern recognition system
 A complete pattern recognition system consists
  of :-
 sensor -gathers the observations to be classified
  or described .
 feature extraction mechanism -computes
  numeric or symbolic information from the
  observations .
 classification or description scheme -that does
  the actual job of classifying or describing
  observations, relying on the extracted features.
Algorithms used by pattern
recognition systems


data                 features
Description task
 The description task transforms data collected from the
  environment into features—
  The description task can involve several different, but
  interrelated, activities:
 Preprocessing:-To modify the data
 Feature extraction:-To generate features
      -- Elementary features
      -- Higher order features
 Feature selection:-To reduce features
The end result of the description task is a
 set of features, commonly called a feature
 vector which constitutes a representation
 of the data.
Classification task

 Uses a classifier to map a feature vector to a
 Such a mapping can be specified by hand or,
  more commonly, a training phase is used to
  induce the mapping from a collection of feature
  vectors known to be representative of the
  various groups among which discrimination is
  being performed (i.e., the training set).
 Once formulated, the mapping can be used to
  assign an identification to each unlabeled
  feature vector subsequently presented to the
Approaches to pattern recognition
There are 2 fundamental approaches to implement
  a pattern recognition system:

1.Statistical (or decision theoretic):-Statistical
  pattern recognition is based on statistical
  characterizations of patterns, assuming that the
  patterns are generated by a probabilistic system.

2. Syntactic (or structural):-Syntactical pattern
   recognition is based on the structural
   interrelationships of features.
Statistical pattern recognition
 It draws from established concepts in statistical
  decision theory to discriminate among data from
  different groups based upon quantitative
  features of the data.
 There are a wide variety of statistical techniques
  that can be used within the description task for
  feature extraction, ranging from simple
  descriptive statistics to complex transformations.
 The quantitative features extracted from each
  object for statistical pattern recognition are
  organized into a fixed length feature vector
  where the meaning associated with each feature
  is determined by its position within the vector.
 The collection of feature vectors generated by
  the description task are passed to the
  classification task.
Syntactic pattern recognition
 Syntactic pattern recognition or structural pattern
  recognition is a form of pattern recognition, where items
  are presented pattern structures which can take into
  account more complex interrelationships between
  features than simple numerical feature vectors used in
  statistical classification.
 It can be used (instead of statistical pattern recognition)
  if there is clear structure in the patterns.
 One way to present such structure is strings of a formal
  language. In this case differences in the structures of the
  classes are encoded as different grammars.
Approaches to pattern recognition
Difference between statistical and
structural pattern recognition
                 statistical               structural
foundation       Statistical decision      Human perception and
                 theory                    cognition

Description      Quantative features       Morphological primitives
                 Fixed no. of features     Variable number of
                 Ignores feature           primitives
                 relationships             Captures primitive
                 Semantics from feature    relationships
                 position                  Semantics from primitive

Classification   Statistical classifiers   Parsing with syntactic
Neural networks pattern recognition

 An “Artificial Neural Network" (ANN), is a
  mathematical model or computational model
  based on biological neural networks. It consists
  of an interconnected group of artificial neurons
  and processes information using a connectionist
  approach to computation.
 In more practical terms neural networks are non-
  linear statistical data modeling tools. They can
  be used to model complex relationships
  between inputs and outputs or to find patterns in
Neural networks pattern recognition

 Classification is based on the response of a network of
  processing units(neurons) to an input stimuli (pattern).
 “Knowledge” is stored in the connectivity and strength
  of the synaptic weights.
 NeurPR is a trainable, non-algorithmic, black-box
 NeurPR is very attractive since
      -it requires minimum a priori knowledge
      -with enough layers and neurons, an ANN can create
       any complex decision region
   Industrial Applications:-
          – Character recognition
          – Process control
          – Signature analysis
          – Speech analysis
   Biometrics:-
          -Face recognition, verification, retrieval
          -Finger prints recognition
          -Speech recognition
   Medical Applications:-
          –X rays
          – Genetic studies
   Government Applications:-
          – Smog detection and measurement
          – Traffic analysis and control
   Military Applications:-
          – Sonar detection and classification
          – Automatic target recognition
 Pattern recognition techniques find applications in
  many areas: machine learning, statistics,
  mathematics, computer science, biology, etc.
 There are many sub-problems in the design process.
 Many of these problems can indeed be solved.
 More complex learning, searching and optimization
  algorithms are developed with advances in computer
 IEEE transactions on Neural Networks.
 Pattern Recognition and Machine Learning.