JAIN ET AL.:STATISTICAL PATTERN RECOGNITION:A REVIEW TABLE 1 Examples of Pattern Recognition Applications Problem Domain Application Input Pattern Pattern Classes Bioinformatics Sequence analysis DNA/Protein sequence Known types of genes/ patterns Data mining Searching for Points in multi- Compact and well- meaningful patterns dimensional space separated clusters Document Internet search Text document Semantic categories classification (e.g.,business,sports, etc.) Document image Reading machine for Document image Alphanumeric analysis the blind characters,words Industrial automation Printed circuit board Intensity or range Defective /non-defective inspection image nature of product Multimedia database Internet search Video clip Video genres (e.g., retrieval action,dialogue,etc.) Biometric recognition Personal identification Face,iris, Authorized users for fingerprint access control Remote sensing Forecasting crop yield Multispectral image Land use categories, growth pattern of crops Speech recognition Telephone directory Speech waveform Spoken words enquiry without operator assistance various physical attributes such as face and fingerprints). well-defined and sufficiently constrained recognition pro- Picard [125]has identified a novel application of pattern blem (small intraclass variations and large interclass recognition,called affective computing which will give a variations)will lead to a compact pattern representation computer the ability to recognize and express emotions,to and a simple decision making strategy.Learning from a set respond intelligently to human emotion,and to employ of examples (training set)is an important and desired mechanisms of emotion that contribute to rational decision attribute of most pattern recognition systems.The four best making.A common characteristic of a number of these known approaches for pattern recognition are:1)template applications is that the available features(typically,in the matching,2)statistical classification,3)syntactic or struc- thousands)are not usually suggested by domain experts, tural matching,and 4)neural networks.These models are but must be extracted and optimized by data-driven not necessarily independent and sometimes the same procedures. pattern recognition method exists with different interpreta- The rapidly growing and available computing power, tions.Attempts have been made to design hybrid systems while enabling faster processing of huge data sets,has also involving multiple models [57].A brief description and facilitated the use of elaborate and diverse methods for data comparison of these approaches is given below and analysis and classification.At the same time,demands on summarized in Table 2. automatic pattern recognition systems are rising enor- 1.2 Template Matching mously due to the availability of large databases and One of the simplest and earliest approaches to pattern stringent performance requirements(speed,accuracy,and recognition is based on template matching.Matching is a cost).In many of the emerging applications,it is clear that no single approach for classification is "optimal"and that generic operation in pattern recognition which is used to determine the similarity between two entities (points, multiple methods and approaches have to be used. curves,or shapes)of the same type.In template matching, Consequently,combining several sensing modalities and a template (typically,a 2D shape)or a prototype of the classifiers is now a commonly used practice in pattern pattern to be recognized is available.The pattern to be recognition. recognized is matched against the stored template while The design of a pattern recognition system essentially taking into account all allowable pose (translation and involves the following three aspects:1)data acquisition and rotation)and scale changes.The similarity measure,often a preprocessing,2)data representation,and 3)decision correlation,may be optimized based on the available making.The problem domain dictates the choice of training set.Often,the template itself is learned from the sensor(s),preprocessing technique,representation scheme, training set.Template matching is computationally de- and the decision making model.It is generally agreed that a manding,but the availability of faster processors has nowvarious physical attributes such as face and fingerprints). Picard [125] has identified a novel application of pattern recognition, called affective computing which will give a computer the ability to recognize and express emotions, to respond intelligently to human emotion, and to employ mechanisms of emotion that contribute to rational decision making. A common characteristic of a number of these applications is that the available features (typically, in the thousands) are not usually suggested by domain experts, but must be extracted and optimized by data-driven procedures. The rapidly growing and available computing power, while enabling faster processing of huge data sets, has also facilitated the use of elaborate and diverse methods for data analysis and classification. At the same time, demands on automatic pattern recognition systems are rising enor￾mously due to the availability of large databases and stringent performance requirements (speed, accuracy, and cost). In many of the emerging applications, it is clear that no single approach for classification is ªoptimalº and that multiple methods and approaches have to be used. Consequently, combining several sensing modalities and classifiers is now a commonly used practice in pattern recognition. The design of a pattern recognition system essentially involves the following three aspects: 1) data acquisition and preprocessing, 2) data representation, and 3) decision making. The problem domain dictates the choice of sensor(s), preprocessing technique, representation scheme, and the decision making model. It is generally agreed that a well-defined and sufficiently constrained recognition pro￾blem (small intraclass variations and large interclass variations) will lead to a compact pattern representation and a simple decision making strategy. Learning from a set of examples (training set) is an important and desired attribute of most pattern recognition systems. The four best known approaches for pattern recognition are: 1) template matching, 2) statistical classification, 3) syntactic or struc￾tural matching, and 4) neural networks. These models are not necessarily independent and sometimes the same pattern recognition method exists with different interpreta￾tions. Attempts have been made to design hybrid systems involving multiple models [57]. A brief description and comparison of these approaches is given below and summarized in Table 2. 1.2 Template Matching One of the simplest and earliest approaches to pattern recognition is based on template matching. Matching is a generic operation in pattern recognition which is used to determine the similarity between two entities (points, curves, or shapes) of the same type. In template matching, a template (typically, a 2D shape) or a prototype of the pattern to be recognized is available. The pattern to be recognized is matched against the stored template while taking into account all allowable pose (translation and rotation) and scale changes. The similarity measure, often a correlation, may be optimized based on the available training set. Often, the template itself is learned from the training set. Template matching is computationally de￾manding, but the availability of faster processors has now JAIN ET AL.: STATISTICAL PATTERN RECOGNITION: A REVIEW 5 TABLE 1 Examples of Pattern Recognition Applications