Pattern Recognition literacy

Pattern Recognition literacy

Post from China's artificial intelligence network

Pattern Recognition: process and analyze various forms of information that characterize things or phenomena (numerical values, text and logical relationships, the process of describing, identifying, classifying, and interpreting things or phenomena is an important part of information science and artificial intelligence.

English "pattern" originated from the French "Patron", originally refers to the ideal person who can be used as a model, or to imitate the perfect copy of the sample. In pattern recognition, "pattern" has a wider significance. When observing things or phenomena, people often need to look for the similarities or differences between things and other things or phenomena, and form incomplete things or phenomena into a category based on a certain purpose. Character Recognition is a typical example. For example, the Chinese character "medium" can be written in a variety of ways, but all belong to the same category. More importantly, even if you have never seen the specific Writing Method of a "medium", you can divide it into the "medium" category. When people are walking on the road, they constantly judge whether they can reach their destination based on the surrounding scenes, this is actually constantly making "correct" and "Incorrect" Classification judgments. This kind of thinking ability of the human brain constitutes the concept of "model. In the above example, a pattern is inseparable from the concept of a category (SET), as long as you know a limited number of things or phenomena in this set, it can recognize any number of things or phenomena in this set. In order to emphasize that we can extract the whole from specific things or phenomena, we call individual things or phenomena a "pattern" and generally a category or category. Some scholars also think that the entire category should be called a model. Such a model is an abstract concept, such as "housing", "Railway", and "Popular Music, A sample of a specific object such as the Great Hall of the People called a "House. Different meanings of such rankings are easy to find out from context.

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. We refer to pattern recognition mainly for speech waveforms, seismic waves, ECG, EEG, pictures, texts, symbols, three objects and scenes, as well as various physical, chemical, and biological sensors. classifies and identifies the specific pattern of an object for measurement.

Pattern recognition is mainly focused on two aspects: How graduate objects (including people) perceive objects, fall into the category of cognitive science, and 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 worked hard for decades through mathematicians, informatics experts, and computer science, we have achieved systematic research results.

Early Research on Computer pattern recognition focuses on model establishment. At the end of 1950s, F. rosenblatt proposed a simplified mathematical model for human brain recognition-perception machine, initially enabling training of the recognition system through various samples of a given category, the system was able to correctly classify models of other unknown categories after learning. In 1960s, the problem was quickly identified using the statistical Decision Theory to solve the problem, A series of monographs on Statistical Pattern Recognition Theories and Methods were published around 1970s. In 1962, R. narasimahan proposed a syntactic recognition method based on the primitive relationship. Fu jingsun worked effectively in this field and formed a systematic theory of syntactic pattern recognition. 1980s J. j. it profoundly reveals the Lenovo storage and computing capabilities of the artificial neural network, and proposes a new way for pattern recognition technology. In just a few years, it has achieved remarkable results in many aspects, thus forming a new discipline direction of the artificial neural network method for pattern recognition.

A computer pattern recognition system consists of three parts: data collection, data processing, and classification decision-making or model matching. Any pattern recognition method must first convert various physical variables of the studied object to a set of numerical values or symbols (strings) acceptable to the computer through various sensors. Traditionally, the space composed of such numbers or symbols (strings) is the pattern space. To extract valid information from these numbers or symbols (strings), it must be processed, including noise elimination, eliminate irrelevant signals and computation of features closely related to the nature of objects and the recognition methods used (for example, characterization of the shape, perimeter, area, and so on) and necessary transformations (such as the fast Fourier transformation to obtain the signal power spectrum. Then, the feature space of the mode is selected through feature selection and extraction or the element. In the future, pattern classification or model matching will be performed based on the feature space. The output of the system, the type of the object, or the model number that is most similar to the object in the model database. For different application purposes, the content of these three parts can be very different, especially in data processing and recognition. In order to improve the reliability of the recognition results, it is often necessary to add a knowledge base (rules) in order to correct possible errors, or greatly reduce the search space of the pattern to be identified in the model library by introducing restrictions to reduce the matching calculation workload. In some specific applications, for example, machine vision, in addition to providing the object to be recognized, it is also necessary to determine the position and posture of the object to guide the robot's work.

Pattern recognition has been successfully applied to weather forecasting, satellite aerial image interpretation, industrial product detection, character recognition, speech recognition, fingerprint recognition, medical image analysis, and many other aspects. All these applications are closely related to the nature of the problem, and have not yet developed into a unified and effective theory that can be applied to all pattern recognition. One common view is that there is no single model used for all pattern recognition problems and a single technology used to solve the recognition problem. We now have a tool bag, what we need to do is to combine the statistical and syntactic (structure) Recognition Methods with specific problems to combine the Statistical Pattern Recognition or syntactic pattern recognition with the heuristic search in artificial intelligence, combining artificial neural network with a variety of technical and artificial intelligence expert systems and uncertain methods, we can gain an in-depth understanding of the efficacy and application possibilities of various tools and learn from each other, creates a new situation for pattern recognition applications.