A non-parameter Bayesian classifier based on Kernel Density Estimation (KDE)is presented for face recognition, which can be regarded as a weighted Nearest Neighbor (NN)classifier in formation. The class conditional de...A non-parameter Bayesian classifier based on Kernel Density Estimation (KDE)is presented for face recognition, which can be regarded as a weighted Nearest Neighbor (NN)classifier in formation. The class conditional density is estimated by KDE and the bandwidthof the kernel function is estimated by Expectation Maximum (EM) algorithm. Two subspaceanalysis methods-linear Principal Component Analysis (PCA) and Kernel-based PCA (KPCA)are respectively used to extract features, and the proposed method is compared with ProbabilisticReasoning Models (PRM), Nearest Center (NC) and NN classifiers which are widely used in facerecognition systems. The experiments are performed on two benchmarks and the experimentalresults show that the KDE outperforms PRM, NC and NN classifiers.展开更多
音频分类是音频信号处理中一项重要的预处理工作。该文描述了一种基于能量的分类方法,将音频信号分为语音和音乐2种类型。分类的过程分为3个阶段,首先计算优化低能量率MLER(modified low energy ratio)作为特征,然后利用初级分类器得到...音频分类是音频信号处理中一项重要的预处理工作。该文描述了一种基于能量的分类方法,将音频信号分为语音和音乐2种类型。分类的过程分为3个阶段,首先计算优化低能量率MLER(modified low energy ratio)作为特征,然后利用初级分类器得到初步分类的结果,最后利用音频类别的前后相关性,使用上下文分类器修正初始分类得到最终分类的结果。该文重点对MLER中参数的合理选取范围进行了讨论,并对传统的初始分类器作了改进,用非参数分类器和参数分类器代替原有的Bayes硬判决的方法,避免了由于门限选择不当所带来的分类错误。实验表明,使用参数分类器时,对纯语音和纯音乐分类效果很好,正确率达99%以上。展开更多
A neural network integrated classifier(NNIC) designed with a new modulation recognition algorithm based on the decision-making tree is proposed in this paper.Firstly,instantaneous parameters are extracted in the time ...A neural network integrated classifier(NNIC) designed with a new modulation recognition algorithm based on the decision-making tree is proposed in this paper.Firstly,instantaneous parameters are extracted in the time domain by the coordinated rotation digital computer(CORDIC) algorithm based on the extended convergence domain and feature parameters of frequency spectrum and power spectrum are extracted by the time-frequency analysis method.All pattern identification parameters are calculated under the I/Q orthogonal two-channel structure,and constructed into the feature vector set.Next,the classifier is designed according to the modulation pattern and recognition performance of the feature parameter set,the optimum threshold is selected for each feature parameter based on the decision-making mechanism in a single classifier,multi-source information fusion and modulation recognition are realized based on feature parameter judge process in the NNIC.Simulation results show NNIC is competent for all modulation recognitions,8 kinds of digital modulated signals are effectively identified,which shows the recognition rate and anti-interference capability at low SNR are improved greatly,the overall recognition rate can reach 100%when SNR is12dB.展开更多
文摘针对入侵检测的特征和分类器参数选择问题,采用极限学习机ELM(extreme learning machine)进行构建分类器,提出一种蝙蝠算法(BA)联合选择特征和分类器参数的网络入侵检测模型(BA-ELM)。首先将特征子集和极限学习机参数编码成蝙蝠个体,以入侵检测准确率和特征数加权组成个体适应度函数;然后通过个体和群体更新的规则引导蝙蝠向最优解飞行,从而找到最优的子特征集和极限学习机参数;最后建立最优的入侵检测模型,并通KDD CUP 99数据集进行仿真性能分析。结果表明,BA-ELM较好地解决了入侵检测特征选择与分类器参数不匹配难题,提高了网络入侵检测率和检测效率,更加适合于网络入侵检测的实时要求。
基金National "863" project (2001AA114140) the National Natural Science Foundation of China (60135020).
文摘A non-parameter Bayesian classifier based on Kernel Density Estimation (KDE)is presented for face recognition, which can be regarded as a weighted Nearest Neighbor (NN)classifier in formation. The class conditional density is estimated by KDE and the bandwidthof the kernel function is estimated by Expectation Maximum (EM) algorithm. Two subspaceanalysis methods-linear Principal Component Analysis (PCA) and Kernel-based PCA (KPCA)are respectively used to extract features, and the proposed method is compared with ProbabilisticReasoning Models (PRM), Nearest Center (NC) and NN classifiers which are widely used in facerecognition systems. The experiments are performed on two benchmarks and the experimentalresults show that the KDE outperforms PRM, NC and NN classifiers.
文摘音频分类是音频信号处理中一项重要的预处理工作。该文描述了一种基于能量的分类方法,将音频信号分为语音和音乐2种类型。分类的过程分为3个阶段,首先计算优化低能量率MLER(modified low energy ratio)作为特征,然后利用初级分类器得到初步分类的结果,最后利用音频类别的前后相关性,使用上下文分类器修正初始分类得到最终分类的结果。该文重点对MLER中参数的合理选取范围进行了讨论,并对传统的初始分类器作了改进,用非参数分类器和参数分类器代替原有的Bayes硬判决的方法,避免了由于门限选择不当所带来的分类错误。实验表明,使用参数分类器时,对纯语音和纯音乐分类效果很好,正确率达99%以上。
基金Supported by the National Natural Science Foundation of China(No.61001049)Key Laboratory of Computer Architecture Opening Topic Fund Subsidization(CARCH201103)Beijing Natural Science Foundation(No.Z2002012201101)
文摘A neural network integrated classifier(NNIC) designed with a new modulation recognition algorithm based on the decision-making tree is proposed in this paper.Firstly,instantaneous parameters are extracted in the time domain by the coordinated rotation digital computer(CORDIC) algorithm based on the extended convergence domain and feature parameters of frequency spectrum and power spectrum are extracted by the time-frequency analysis method.All pattern identification parameters are calculated under the I/Q orthogonal two-channel structure,and constructed into the feature vector set.Next,the classifier is designed according to the modulation pattern and recognition performance of the feature parameter set,the optimum threshold is selected for each feature parameter based on the decision-making mechanism in a single classifier,multi-source information fusion and modulation recognition are realized based on feature parameter judge process in the NNIC.Simulation results show NNIC is competent for all modulation recognitions,8 kinds of digital modulated signals are effectively identified,which shows the recognition rate and anti-interference capability at low SNR are improved greatly,the overall recognition rate can reach 100%when SNR is12dB.