Pattern Recognition Overview

Pattern recognition is also called pattern classification. From the perspective of the nature of the problem and the method for solving the problem, pattern recognition is classified into supervised classification (supervised classification) and unsupervised classification. Pattern recognition is a basic intelligence of human beings. In daily life, people often perform pattern recognition ". With the emergence of computers in 1940s and the rise of artificial intelligence in 1950s, people certainly hope to use computers to replace or expand part of human mental work. (Computer) pattern recognition developed rapidly in the early 1960s s and has become a new discipline.
Pattern Recognition refers to the processing and analysis of various forms of information (numerical value, text and logical relationship) that characterize things or phenomena, the process of describing, identifying, classifying, and interpreting things or phenomena is an important part of information science and artificial intelligence.
The mode can also be divided into two forms: abstract and concrete. The former, such as consciousness, thoughts, and arguments, belongs to the scope of concept Recognition Research and is another branch of AI research. The pattern recognition we refer to is mainly to classify and identify the specific pattern of the objects such as voice waveforms, seismic waves, ECG, EEG, pictures, photos, texts, symbols, and biological sensors.
Pattern recognition is mainly focused on two aspects: first, how graduate objects (including people) perceive objects belongs to the scope of cognitive science, and second, under the given task, how to use a computer to implement the theory and method of pattern recognition. The former is the content of research by scientists, psychologists, bioscientists and neuroscientists. The latter has made systematic research achievements through the efforts of mathematicians, informatics experts and computer scientists in recent decades.
The Application computer identifies and classifies a group of events or processes. The recognized events or processes can be specific objects such as words, sounds, and images, or abstract objects such as States and degrees. These objects are different from digital information, called pattern information.
The number of categories of pattern recognition is determined by a specific recognition problem. Sometimes, the actual number of classes cannot be known at the beginning, and the system needs to identify the objects to be identified after repeated observation.
Pattern recognition is related to statistics, psychology, linguistics, computer science, biology, and control theory. It is correlated with the research of artificial intelligence and image processing. For example, an adaptive or self-organized pattern recognition system includes the Learning Mechanism of artificial intelligence. The scene understanding and natural language understanding of Artificial Intelligence also contain pattern recognition problems. Another example is the application of image processing technology in the pre-processing and feature extraction processes in pattern recognition. Image Analysis in image processing also applies pattern recognition technology.
I. Pattern Recognition Methods
1. Theoretical decision-making methods
Also known as the statistical method, it is a method that is developed earlier and more mature. The identified object is first digitalized and transformed into digital information suitable for computer processing. A mode is often expressed with a large amount of information. Many pattern recognition systems are preprocessed after digitization to remove mixed interference information and reduce some deformation and distortion. Then, we extract a set of features from the input mode after digitization or preprocessing. The so-called feature is a selected measurement. It remains unchanged or almost unchanged for general deformation and distortion, and only contains as little redundant information as possible. During feature extraction, the input mode is mapped from the object space to the feature space. In this case, the mode can be represented by a vertex or a feature vector in the feature space. This ing not only compresses the amount of information, but also facilitates classification. In decision-making theory, feature extraction plays an important role, but there is no general theoretical guidance. We can only determine which feature to select by analyzing specific recognition objects. Features can be classified after extraction, that is, from feature space to decision space. For this reason, the identification function is introduced, and the feature vector is used to calculate the corresponding identification function values, and classification is implemented by comparing the identification function values.
2. Syntax
Also known as structural or linguistic methods. The basic idea is to describe a mode as a simple combination of sub-modes. The sub-mode can also be described as a simpler combination of sub-modes, and finally get a tree structure description, the simplest sub-mode at the underlying layer is called a schema primitive. Selecting a primitive in the syntax method is equivalent to selecting a feature in the decision theory method. It is usually required that the selected primitive can provide a compact description of the schema to reflect its structural relationship, and be easily extracted using non-syntactic methods. Obviously, the primitive itself should not contain important structure information. A mode is described by a group of elements and their combinations. It is called a mode description statement. This is equivalent to a combination of sentences and phrases in a language, with the same combination of words and characters. A rule that combines elements into a pattern, which is specified by the so-called syntax. Once the primitive is identified, the recognition process can be performed through syntactic analysis, that is, to analyze whether the given schema statement conforms to the specified syntax. If a syntax is satisfied, it is classified into the class.
The choice of pattern recognition method depends on the nature of the problem. If the identified object is extremely complex and contains rich structure information, the syntax method is generally used. The identified object is not very complex or does not contain obvious structure information, and the decision-making theory method is generally used. The two methods cannot be completely separated. In the syntax method, the elements are extracted using the decision theory method. In applications, the two methods are combined to apply to different layers, and the results are often better.
II. Application of Pattern Recognition
Pattern recognition can be used in text and speech recognition, remote sensing and medical diagnosis.
① Text Recognition
Chinese characters have a history of thousands of years and are the most widely used Chinese characters in the world. They have an indelible honor for the formation and development of the splendid Chinese culture. Therefore, with the increasing popularity of information and computer technologies, how to easily and quickly input texts into computers has become an important bottleneck affecting the efficiency of man-machine interfaces, it is also related to whether computers can be used widely. Currently, Chinese character input can be divided into manual keyboard input and automatic machine recognition input. Manual input is slow and labor intensive. Automatic Input is classified into Chinese character recognition input and speech recognition input. In terms of the difficulty of recognition technology, the difficulty of handwritten recognition is higher than that of printed recognition. In handwritten recognition, the difficulty of Offline Handwritten recognition is far greater than that of connected handwritten recognition. So far, in addition to the practical application of Offline handwritten numbers, Offline Handwritten recognition of Chinese characters and other words is still in the laboratory stage.
② Speech Recognition
Speech recognition technology involves the following fields: signal processing, pattern recognition, probability theory and information theory, vocal mechanism and auditory mechanism, and artificial intelligence. In recent years, in the field of biological recognition technology, voiceprint recognition technology has attracted worldwide attention for its unique convenience, economy, and accuracy, and increasingly become an important and popular security verification method in people's daily life and work. Furthermore, Using Genetic Algorithms to train the Continuous Hidden Markov Model for Speech Recognition has become a mainstream technology in speech recognition. This method is fast in speech recognition and has a high recognition rate. 2.3 Fingerprint Recognition
The uneven skin on the inside of our palms, fingers, feet, and toes forms various patterns. The pattern, breakpoint, and intersection of these skins are different and unique. With this uniqueness, you can compare a person's fingerprint with his fingerprint to verify his real identity. General fingerprints are classified into the following major categories: Left Loop, Right Loop, twin loop, whorl, arch, and tented arch. This way, the fingerprints of each person can be classified and retrieved separately. Fingerprint recognition can be divided into several major steps: preprocessing, feature selection, and pattern classification.
③ Remote Sensing
Remote sensing image recognition has been widely used in crop estimation, resource survey, weather forecasting and military reconnaissance.
④ Medical Diagnosis
Pattern recognition has achieved remarkable results in cancer cell detection, X-ray photo analysis, blood tests, chromosome analysis, ECG diagnosis, and EEG diagnosis.
Iii. Statistical Pattern Recognition
The basic principle of statistical pattern recognition is that samples with similarity are close to each other in the pattern space and form a "group", that is, "together by things ". The analysis method is based on the feature vectors xi = (xi1, xi2 ,..., Xid) T (I = 1, 2 ,..., N), classify a given pattern into Class C ω 1, ω 2 ,..., In ω C, the classification is determined based on the distance function between the modes. T indicates transpose, N indicates the number of sample points, and D indicates the number of sample features.
The main methods for Statistical Pattern Recognition include: discriminant function, k-nn, nonlinear ing, feature analysis, and main factor analysis.
In statistical pattern recognition, Bayesian decision-making rules theoretically solve the problem of optimal classifier design. However, the implementation of Bayesian decision-making rules must first solve the more difficult issue of probability density estimation. BP Neural Networks learn from observation data (training samples) directly, which is a simpler and more effective method and has been widely used. However, it is a heuristic technique, lack of a solid theoretical foundation for specified engineering practices. The breakthrough achievements made by the Institute of statistical inference theory lead to the establishment of the modern statistical learning theory-VC Theory. This theory not only answered the theoretical problems in the artificial neural network on a strict mathematical basis, in addition, a new learning method, support vector machine, is derived.
Iv. almost unlimited development potential of Pattern Recognition Technology
Pattern recognition technology is the basic technology of artificial intelligence. The 21st century is the century of intelligence, information, computing, and network. In this century, which features digital computing, as the basic discipline of artificial intelligence technology, pattern recognition technology will surely gain a huge space for development. Internationally, various authoritative research institutions and companies have begun to pay attention to pattern recognition technology as the company's strategic R & D focus.
1. Speech Recognition Technology
Speech recognition technology is gradually becoming a key technology in information technology for human-machine interfaces. The application of speech technology has become a competitive emerging high-tech industry. China's Internet center market forecast: in the next five years, the Chinese speech technology field will have a market capacity of more than 40 billion RMB, and then grow at an annual rate of more than 30%.
2. Biological Authentication Technology
The development of biological authentication technology, the most important security authentication technology in this century, is the trend of the times. People are willing to forget all their passwords, discard all their magnetic cards, and identify and keep them confidential based on their uniqueness. IDC prediction: as the inevitable future development direction, biometric identification technology, the core technology of mobile e-commerce, will reach a market scale of $10 billion in the next 10 years.
3. Digital Watermarking Technology
The digital watermarking technology that began to develop internationally since 1990s is the most promising and advantageous digital media copyright protection technology. IDC predicts that the capacity of digital watermarking technology in the global market will exceed $8 billion in the next five years.
V. Jieyu
Since its development in 1920s, pattern recognition has never been applied to all pattern recognition problems, what we have now is a tool bag. What we need to do is to combine the statistical and syntactic recognition with specific problems, combine Statistical Pattern Recognition or Syntactic Pattern Recognition with heuristic search in artificial intelligence, and combine statistical pattern recognition or Syntactic Pattern Recognition with machine learning of SVM, the artificial neural network is combined with various existing technologies, as well as expert systems in artificial intelligence, and Uncertain Reasoning Methods to gain an in-depth understanding of the efficacy and possibility of various tools and to learn from each other, creates a new situation for pattern recognition applications.
There are various theoretical interpretations of the ability to recognize two-dimensional patterns. The template says that every pattern we know has a corresponding template or thumbnail copy in long memory. Pattern recognition is the most suitable template for visual stimulation. In terms of features, visual stimulation is composed of various features. pattern recognition is a model feature that presents stimulus and is stored in long-term memory. Feature description explains some bottom-up processes in pattern recognition, but it does not emphasize environment-based information and top-down processing. The structure-based description theory may be more appropriate than the template or feature theory.