patternrecognition problems occur in many domains and consequently well performing classification algorithms are highly sought after. In my talk, I will focus on a particular group of classifiers that generate rule b...
详细信息
patternrecognition problems occur in many domains and consequently well performing classification algorithms are highly sought after. In my talk, I will focus on a particular group of classifiers that generate rule bases consisting of a set of fuzzy if‐then rules. The antecedent part of a rule characterises the input features using fuzzy sets, while the consequent part is established in a learning stage. Starting with a basic fuzzy classifier I will show that, through a simple modification, it can be turned into a cost sensitive classification method, that classification performance can be improved through an error correction learning approach, and that a small yet powerful rule base can be generated through an optimisation approach based on genetic algorithms.
Word level Script and language identification is a process of separating the script and language of each word present in a printed or handwritten multi-script document. It is an essential part of a multi-lingual Optic...
详细信息
Word level Script and language identification is a process of separating the script and language of each word present in a printed or handwritten multi-script document. It is an essential part of a multi-lingual Optical Character Recognizer (OCR). Most of the OCRs are solely designed for a single script. So it can't convert a document which is written in more than one script. This paper explained a system, which automatically separate the script of an unconstrained handwritten document mix with three Indian scripts and Roman script English up to word level. The process starts from extracting text-lines from the document and then separates the words from the text-line using projection profiles. 15 connected component and morphological features and 32 Gabor filter features is extracted from one word to form a feature set. Total of 15699 words are separated from 215 documents for training and 4856 words from 64 documents for testing. Three different classifiers, Support Vector Machine (SVM), Multilayer Perceptron (MLP), and K-Nearest Neighbors (KNN) classifiers are used for testing the discriminating power of the feature set. MLP classifier outperform over all others with cross validation accuracy of 81.69% across four scripts.
This paper proposes pattern analysis of golf swing from data collection *** system uses motion sensors of 9DOF Razor IMU to collect the data by attaching Razor IMUs to 3 positions of human body:upper back,lower back,a...
详细信息
This paper proposes pattern analysis of golf swing from data collection *** system uses motion sensors of 9DOF Razor IMU to collect the data by attaching Razor IMUs to 3 positions of human body:upper back,lower back,and right *** sensors can detect the patterns of body movement in golfing from the beginning to the *** types of data collected are acceleration and *** output of the data from the sensors can indicate the appropriate patterns of golf *** experimental result data from the sensors are illustrated in graphs showing each step of movement patterns in golf *** movement patterns can be beneficial for golf swing trainers in ***,they can help doctors to analyze the cause of golfers’injury.
The aim of this ongoing research project is to reach a patternrecognition process(design system) that can turn architecture to music, and turn music to architecture.. In both operations, style and character will be a...
详细信息
The aim of this ongoing research project is to reach a patternrecognition process(design system) that can turn architecture to music, and turn music to architecture.. In both operations, style and character will be an essential issue in the design of this process system. Based on the artistic relationship between architecture and music, we shall try to settle a mechanism for a patternrecognition audio-visual system that can that can turn a certain style of architecture to a collection of melodies that truly reflects the character of that architecture. This system can also perform the the inverted process, that is turning a certain musical style to a collection of shapes and forms. In this regard it is important to point out to the interrelationship between the architectural design process and writing a musical note. A musician composing a melody works from above, that is from the whole to the parts. A composer does not usually put notes together in order to get some melody;he envisages the character of a melody in a statu nascendi and proceeds from above as he tries to concretize it in all its part. In this respect, the articulation of the whole and the important junctures of the parts, the beginning(s), or the end(s), the composer accepted the attitude of his contemporaries, and could therefore integrate principles of architecture such as the golden section and the Pythagorean Theorem, along with principles of acoustic harmony.
In this paper,we improved the method of minimum squared error for robust classification by altering its classification *** minimum squared error method,is one of the methods to minimize the sum of the squared error be...
详细信息
In this paper,we improved the method of minimum squared error for robust classification by altering its classification *** minimum squared error method,is one of the methods to minimize the sum of the squared error between the output of the linear function and the desired output,which first obtains the mapping that can best transform the training sample into its class label and then exploits the obtained mapping to predict the class label of the test ***,the improved method classifies all the test examples by comparing the difference of predicted class labels of test examples and training *** experiment in section 4,we draw a conclusion that the improved MSE(IMSE) method can reduce the classification error and get a better classification result.
In this paper,a novel violent video detection scheme is presented based on visual *** violent segment is detected by using motion angle,motion intensity,shot & explosion and blood *** addition,in order to improve ...
详细信息
In this paper,a novel violent video detection scheme is presented based on visual *** violent segment is detected by using motion angle,motion intensity,shot & explosion and blood *** addition,in order to improve the compute efficiency and time savings Hidden Information SVM is *** result shows that the proposed method is effective in violent video detection.
For the patternrecognition problem, this paper proposes a feature selection method based on complementarity analysis. Analyse the separability of single feature, search the feature combination with the smallest proba...
详细信息
ISBN:
(纸本)9781479928606
For the patternrecognition problem, this paper proposes a feature selection method based on complementarity analysis. Analyse the separability of single feature, search the feature combination with the smallest probability of mixing region and in the mixed region with the greatest separability to reduce the probability of classification error. Compared with other feature selection algorithms, data testing result shows that the feature selection method based on complementarity analysis has a lower error recognition rate than other methods, which has verified the superiority and the advanced nature of the method.
In this paper, we have proposed a framework for representation and recognition of motion events between two or more objects. The framework combines qualitative spatial abstractions with the technique of syntactic patt...
详细信息
ISBN:
(纸本)9781467345286
In this paper, we have proposed a framework for representation and recognition of motion events between two or more objects. The framework combines qualitative spatial abstractions with the technique of syntactic patternrecognition in order to recognize motion patterns in an input stream. The key feature of the framework is recognition of motion events among multiple objects and handling of temporal constraints among events.
pattern structures propose a direct way to knowledge discovery in data with structure, such as logical formulas, graphs, strings, tuples of numerical intervals, etc., by defining closed descriptions and discovery tool...
详细信息
Non-negative matrix factorization and sparse representation models have been successfully applied in high-throughput biological data analysis. In this paper, we propose our versatile sparse matrix factorization (VSMF)...
详细信息
暂无评论