Is there a preferred classifier for operational thematic mapping?

The importance of properly exploiting a classifier's inherent geometric characteristics when developing a classification methodology is emphasized as a prerequisite to achieving near optimal performance when carrying out thematic mapping. When used properly, it is argued that the long-standing maximum likelihood approach and the more recent support vector machine can perform comparably. Both contain the flexibility to segment the spectral domain in such a manner as to match inherent class separations in the data, as do most reasonable classifiers. The choice of which classifier to use in practice is determined largely by preference and related considerations, such as ease of training, multiclass capabilities, and classification cost. © 1980-2012 IEEE.

Biomimetic pattern recognition theory and its applications

Biomimetic pattern recogntion (BPR), which is based on "cognition" instead of "classification", is much closer to the function of human being. The basis of BPR is the Principle of homology-continuity (PHC), which means the difference between two samples of the same class must be gradually changed. The aim of BPR is to find an optimal covering in the feature space, which emphasizes the "similarity" among homologous group members, rather than "division" in traditional pattern recognition. Some applications of BPR are surveyed, in which the results of BPR are much better than the results of Support Vector Machine. A novel neuron model, Hyper sausage neuron (HSN), is shown as a kind of covering units in BPR. The mathematical description of HSN is given and the 2-dimensional discriminant boundary of HSN is shown. In two special cases, in which samples are distributed in a line segment and a circle, both the HSN networks and RBF networks are used for covering. The results show that HSN networks act better than RBF networks in generalization, especially for small sample set, which are consonant with the results of the applications of BPR. And a brief explanation of the HSN networks' advantages in covering general distributed samples is also given.

Mandarin speech emotion recognition based on high dimensional geometry theory

In this paper, a novel approach for mandarin speech emotion recognition, that is mandarin speech emotion recognition based on high dimensional geometry theory, is proposed. The human emotions are classified into 6 archetypal classes: fear, anger, happiness, sadness, surprise and disgust. According to the characteristics of these emotional speech signals, the amplitude, pitch frequency and formant are used as the feature parameters for speech emotion recognition. The new method called high dimensional geometry theory is applied for recognition. Compared with traditional GSVM model, the new method has some advantages. It is noted that this method has significant values for researches and applications henceforth.

基于支持向量机的Internet流量分类研究

准确的网络流量分类是众多网络研究工作的基础,也一直是网络测量领域的研究热点.近年来,利用机器学习方法处理流量分类问题成为了该领域一个新兴的研究方向.在目前研究中应用较多的是朴素贝叶斯(nave Bayes,NB)及其改进算法.这些方法具有实现简单、分类高效的特点.但该方法过分依赖于样本空间的分布,具有内在的不稳定性.因此,提出一种基于支持向量机(support vector machine,SVM)的流量分类方法.该方法利用非线性变换和结构风险最小化(structural risk minimization,SRM)原则将流量分类问题转化为二次寻优问题,具有良好的分类准确率和稳定性.在理论分析的基础上,通过在实际网络流集合上与朴素贝叶斯算法的对比实验,可以看出使用支持向量机方法处理流量分类问题,具有以下3个优势:1)网络流属性不必满足条件独立假设,无须进行属性过滤;2)能够在先验知识相对不足的情况下,仍保持较高的分类准确率;3)不依赖于样本空间的分布,具有较好的分类稳定性.

Efficient HOG human detection

While Histograms of Oriented Gradients (HOG) plus Support Vector Machine (SVM) (HOG+SVM) is the most successful human detection algorithm, it is time-consuming. This paper proposes two ways to deal with this problem. One way is to reuse the features in blocks to construct the HOG features for intersecting detection windows. Another way is to utilize sub-cell based interpolation to efficiently compute the HOG features for each block. The combination of the two ways results in significant increase in detecting humans-more than five times better. To evaluate the proposed method, we have established a top-view human database. Experimental results on the top-view database and the well-known INRIA data set have demonstrated the effectiveness and efficiency of the proposed method. (C) 2010 Elsevier B.V. All rights reserved.

基于独立成分分析和模糊支持向量机的人脸识别方法

成分分析(PCA)只能从2阶上去消除数据的相关性，传统支持向量机在解决多类问题时会出现分类的盲区问题，针对这两种情况，首先采用独立成分分析(ICA)方法解决了高阶上的数据相关性问题；同时在传统支持向量机中引入模糊隶属度函数，用模糊支持向量机解决了传统支持向量机在多类数据识别中的盲区问题。通过实验证明了该方法在人脸识别率上取得了显著提高。

滑坡泥石流信息遥感获取与空间预测

On the issue of geological hazard evaluation（GHE）, taking remote sensing and GIS systems as experimental environment, assisting with some programming development, this thesis combines multi-knowledges of geo-hazard mechanism, statistic learning, remote sensing (RS), high-spectral recognition, spatial analysis, digital photogrammetry as well as mineralogy, and selects geo-hazard samples from Hong Kong and Three Parallel River region as experimental data, to study two kinds of core questions of GHE, geo-hazard information acquiring and evaluation model. In the aspect of landslide information acquiring by RS, three detailed topics are presented, image enhance for visual interpretation, automatic recognition of landslide as well as quantitative mineral mapping. As to the evaluation model, the latest and powerful data mining method, support vector machine (SVM), is introduced to GHE field, and a serious of comparing experiments are carried out to verify its feasibility and efficiency. Furthermore, this paper proposes a method to forecast the distribution of landslides if rainfall in future is known baseing on historical rainfall and corresponding landslide susceptibility map. The details are as following: (a) Remote sensing image enhancing methods for geo-hazard visual interpretation. The effect of visual interpretation is determined by RS data and image enhancing method, for which the most effective and regular technique is image merge between high-spatial image and multi-spectral image, but there are few researches concerning the merging methods of geo-hazard recognition. By the comparing experimental of six mainstream merging methods and combination of different remote sensing data source, this thesis presents merits of each method ,and qualitatively analyzes the effect of spatial resolution, spectral resolution and time phase on merging image. (b) Automatic recognition of shallow landslide by RS image. The inventory of landslide is the base of landslide forecast and landslide study. If persistent collecting of landslide events, updating the geo-hazard inventory in time, and promoting prediction model incessantly, the accuracy of forecast would be boosted step by step. RS technique is a feasible method to obtain landslide information, which is determined by the feature of geo-hazard distribution. An automatic hierarchical approach is proposed to identify shallow landslides in vegetable region by the combination of multi-spectral RS imagery and DEM derivatives, and the experiment is also drilled to inspect its efficiency. (c) Hazard-causing factors obtaining. Accurate environmental factors are the key to analyze and predict the risk of regional geological hazard. As to predict huge debris flow, the main challenge is still to determine the startup material and its volume in debris flow source region. Exerting the merits of various RS technique, this thesis presents the methods to obtain two important hazard-causing factors, DEM and alteration mineral, and through spatial analysis, finds the relationship between hydrothermal clay alteration minerals and geo-hazards in the arid-hot valleys of Three Parallel Rivers region. (d) Applying support vector machine (SVM) to landslide susceptibility mapping. Introduce the latest and powerful statistical learning theory, SVM, to RGHE. SVM that proved an efficient statistic learning method can deal with two-class and one-class samples, with feature avoiding produce ‘pseudo’ samples. 55 years historical samples in a natural terrain of Hong Kong are used to assess this method, whose susceptibility maps obtained by one-class SVM and two-class SVM are compared to that obtained by logistic regression method. It can conclude that two-class SVM possesses better prediction efficiency than logistic regression and one-class SVM. However, one-class SVM, only requires failed cases, has an advantage over the other two methods as only "failed" case information is usually available in landslide susceptibility mapping. (e) Predicting the distribution of rainfall-induced landslides by time-series analysis. Rainfall is the most dominating factor to bring in landslides. More than 90% losing and casualty by landslides is introduced by rainfall, so predicting landslide sites under certain rainfall is an important geological evaluating issue. With full considering the contribution of stable factors (landslide susceptibility map) and dynamic factors (rainfall), the time-series linear regression analysis between rainfall and landslide risk mapis presented, and experiments based on true samples prove that this method is perfect in natural region of Hong Kong. The following 4 practicable or original findings are obtained: 1) The RS ways to enhance geo-hazards image, automatic recognize shallow landslides, obtain DEM and mineral are studied, and the detailed operating steps are given through examples. The conclusion is practical strongly. 2) The explorative researching about relationship between geo-hazards and alteration mineral in arid-hot valley of Jinshajiang river is presented. Based on standard USGS mineral spectrum, the distribution of hydrothermal alteration mineral is mapped by SAM method. Through statistic analysis between debris flows and hazard-causing factors, the strong correlation between debris flows and clay minerals is found and validated. 3) Applying SVM theory (especially one-class SVM theory) to the landslide susceptibility mapping and system evaluation for its performance is also carried out, which proves that advantages of SVM in this field. 4) Establishing time-serial prediction method for rainfall induced landslide distribution. In a natural study area, the distribution of landslides induced by a storm is predicted successfully under a real maximum 24h rainfall based on the regression between 4 historical storms and corresponding landslides.

社会形势预警的几种建模方法比较

Based on social survey data conducted by local research group in some counties executed in the nearly past five years in China, the author proposed and solved two kernel problems in the field of social situation forecasting: i) How can the attitudes’ data on individual level be integrated with social situation data on macrolevel; ii) How can the powers of forecasting models’ constructed by different statistic methods be compared? Five integrative statistics were applied to the research: 1) algorithm average (MEAN); 2) standard deviation (SD); 3) coefficient variability (CV); 4) mixed secondary moment (M2); 5) Tendency (TD). To solve the former problem, the five statistics were taken to synthesize the individual and mocrolevel data of social situations on the levels of counties’ regions, and form novel integrative datasets, from the basis of which, the latter problem was accomplished by the author: modeling methods such as Multiple Regression Analysis (MRA), Discriminant Analysis (DA) and Support Vector Machine (SVM) were used to construct several forecasting models. Meanwhile, on the dimensions of stepwise vs. enter, short-term vs. long-term forecasting and different integrative (statistic) models, meta-analysis and power analysis were taken to compare the predicting power of each model within and among modeling methods. Finally, it can be concluded from the research of the dissertation: 1) Exactly significant difference exists among different integrative (statistic) models, in which, tendency (TD) integrative models have the highest power, but coefficient variability (CV) ones have the lowest; 2) There is no significant difference of the power between stepwise and enter models as well as short-term and long-term forecasting models; 3) There is significant difference among models constructed by different methods, of which, support vector machine (SVM) has the highest statistic power. This research founded basis in all facets for exploring the optimal forecasting models of social situation’s more deeply, further more, it is the first time methods of meta-analysis and power analysis were immersed into the assessments of such forecasting models.

An alignment based similarity measure for hand detection in cluttered sign language video

Locating hands in sign language video is challenging due to a number of factors. Hand appearance varies widely across signers due to anthropometric variations and varying levels of signer proficiency. Video can be captured under varying illumination, camera resolutions, and levels of scene clutter, e.g., high-res video captured in a studio vs. low-res video gathered by a web cam in a user’s home. Moreover, the signers’ clothing varies, e.g., skin-toned clothing vs. contrasting clothing, short-sleeved vs. long-sleeved shirts, etc. In this work, the hand detection problem is addressed in an appearance matching framework. The Histogram of Oriented Gradient (HOG) based matching score function is reformulated to allow non-rigid alignment between pairs of images to account for hand shape variation. The resulting alignment score is used within a Support Vector Machine hand/not-hand classifier for hand detection. The new matching score function yields improved performance (in ROC area and hand detection rate) over the Vocabulary Guided Pyramid Match Kernel (VGPMK) and the traditional, rigid HOG distance on American Sign Language video gestured by expert signers. The proposed match score function is computationally less expensive (for training and testing), has fewer parameters and is less sensitive to parameter settings than VGPMK. The proposed detector works well on test sequences from an inexpert signer in a non-studio setting with cluttered background.

Learning a Family of Detectors

Object detection and recognition are important problems in computer vision. The challenges of these problems come from the presence of noise, background clutter, large within class variations of the object class and limited training data. In addition, the computational complexity in the recognition process is also a concern in practice. In this thesis, we propose one approach to handle the problem of detecting an object class that exhibits large within-class variations, and a second approach to speed up the classification processes. In the first approach, we show that foreground-background classification (detection) and within-class classification of the foreground class (pose estimation) can be jointly solved with using a multiplicative form of two kernel functions. One kernel measures similarity for foreground-background classification. The other kernel accounts for latent factors that control within-class variation and implicitly enables feature sharing among foreground training samples. For applications where explicit parameterization of the within-class states is unavailable, a nonparametric formulation of the kernel can be constructed with a proper foreground distance/similarity measure. Detector training is accomplished via standard Support Vector Machine learning. The resulting detectors are tuned to specific variations in the foreground class. They also serve to evaluate hypotheses of the foreground state. When the image masks for foreground objects are provided in training, the detectors can also produce object segmentation. Methods for generating a representative sample set of detectors are proposed that can enable efficient detection and tracking. In addition, because individual detectors verify hypotheses of foreground state, they can also be incorporated in a tracking-by-detection frame work to recover foreground state in image sequences. To run the detectors efficiently at the online stage, an input-sensitive speedup strategy is proposed to select the most relevant detectors quickly. The proposed approach is tested on data sets of human hands, vehicles and human faces. On all data sets, the proposed approach achieves improved detection accuracy over the best competing approaches. In the second part of the thesis, we formulate a filter-and-refine scheme to speed up recognition processes. The binary outputs of the weak classifiers in a boosted detector are used to identify a small number of candidate foreground state hypotheses quickly via Hamming distance or weighted Hamming distance. The approach is evaluated in three applications: face recognition on the face recognition grand challenge version 2 data set, hand shape detection and parameter estimation on a hand data set, and vehicle detection and estimation of the view angle on a multi-pose vehicle data set. On all data sets, our approach is at least five times faster than simply evaluating all foreground state hypotheses with virtually no loss in classification accuracy.

Neonatal EEG classification using atomic decomposition

The electroencephalogram (EEG) is an important noninvasive tool used in the neonatal intensive care unit (NICU) for the neurologic evaluation of the sick newborn infant. It provides an excellent assessment of at-risk newborns and formulates a prognosis for long-term neurologic outcome.The automated analysis of neonatal EEG data in the NICU can provide valuable information to the clinician facilitating medical intervention. The aim of this thesis is to develop a system for automatic classification of neonatal EEG which can be mainly divided into two parts: (1) classification of neonatal EEG seizure from nonseizure, and (2) classifying neonatal background EEG into several grades based on the severity of the injury using atomic decomposition. Atomic decomposition techniques use redundant time-frequency dictionaries for sparse signal representations or approximations. The first novel contribution of this thesis is the development of a novel time-frequency dictionary coherent with the neonatal EEG seizure states. This dictionary was able to track the time-varying nature of the EEG signal. It was shown that by using atomic decomposition and the proposed novel dictionary, the neonatal EEG transition from nonseizure to seizure states could be detected efficiently. The second novel contribution of this thesis is the development of a neonatal seizure detection algorithm using several time-frequency features from the proposed novel dictionary. It was shown that the time-frequency features obtained from the atoms in the novel dictionary improved the seizure detection accuracy when compared to that obtained from the raw EEG signal. With the assistance of a supervised multiclass SVM classifier and several timefrequency features, several methods to automatically grade EEG were explored. In summary, the novel techniques proposed in this thesis contribute to the application of advanced signal processing techniques for automatic assessment of neonatal EEG recordings.

Model selection approaches for non-linear system identification: a review

The identification of non-linear systems using only observed finite datasets has become a mature research area over the last two decades. A class of linear-in-the-parameter models with universal approximation capabilities have been intensively studied and widely used due to the availability of many linear-learning algorithms and their inherent convergence conditions. This article presents a systematic overview of basic research on model selection approaches for linear-in-the-parameter models. One of the fundamental problems in non-linear system identification is to find the minimal model with the best model generalisation performance from observational data only. The important concepts in achieving good model generalisation used in various non-linear system-identification algorithms are first reviewed, including Bayesian parameter regularisation and models selective criteria based on the cross validation and experimental design. A significant advance in machine learning has been the development of the support vector machine as a means for identifying kernel models based on the structural risk minimisation principle. The developments on the convex optimisation-based model construction algorithms including the support vector regression algorithms are outlined. Input selection algorithms and on-line system identification algorithms are also included in this review. Finally, some industrial applications of non-linear models are discussed.

SVM Training Phase Reduction Using Dataset Feature Filtering for Malware Detection

N-gram analysis is an approach that investigates the structure of a program using bytes, characters, or text strings. A key issue with N-gram analysis is feature selection amidst the explosion of features that occurs when N is increased. The experiments within this paper represent programs as operational code (opcode) density histograms gained through dynamic analysis. A support vector machine is used to create a reference model, which is used to evaluate two methods of feature reduction, which are 'area of intersect' and 'subspace analysis using eigenvectors.' The findings show that the relationships between features are complex and simple statistics filtering approaches do not provide a viable approach. However, eigenvector subspace analysis produces a suitable filter.

Adaptive Bacteria Colony Picking in Unstructured Environments Using Intensity Histogram and Unascertained LS-SVM Classifier

Features analysis is an important task which can significantly affect the performance of automatic bacteria colony picking. Unstructured environments also affect the automatic colony screening. This paper presents a novel approach for adaptive colony segmentation in unstructured environments by treating the detected peaks of intensity histograms as a morphological feature of images. In order to avoid disturbing peaks, an entropy based mean shift filter is introduced to smooth images as a preprocessing step. The relevance and importance of these features can be determined in an improved support vector machine classifier using unascertained least square estimation. Experimental results show that the proposed unascertained least square support vector machine (ULSSVM) has better recognition accuracy than the other state-of-the-art techniques, and its training process takes less time than most of the traditional approaches presented in this paper.