Pattern recognition using discriminative feature extraction

We propose a new design method, called discriminative feature extraction for practical modular pattern recognizers. A key concept of discriminative feature extraction is the design of an overall recognizer in a manner consistent with recognition error minimization. The utility of the method is demonstrated in a Japanese vowel recognition task     

多尺度局部结构主导二值模式学习图像表示

通过零均值化的微观结构模式二值化(ZMPB)处理,该文提出一种立足于局部图像多尺度结构二值模式提取的图像表示方法。该方法能够表达图像中可能出现的各种具有视觉意义的重要模式结构,同时通过主导二值模式学习模型,可以获得适应于图像数据集的主导特征模式子集,在特征鲁棒性、鉴别力和表达能力上达到优异性能,同时可以有效降低特征编码的维度,提高算法的执行速度。实验结果表明该算法性能优异,具有很强的鉴别能力和鲁棒性,优于传统LBP和GIMMRP方法,和很多最新算法结果相比,也具有竞争优势。     

High-dimensional feature extraction using bit-plane decomposition of local binary patterns for robust face recognition

Transforming an original image into a high-dimensional (HD) feature has been proven to be effective in classifying images. This paper presents a novel feature extraction method utilizing the HD feature space to improve the discriminative ability for face recognition. We observed that the local binary pattern can be decomposed into bit-planes, each of which has scale-specific directional information of the face image. Each bit-plane not only has the inherent local-structure of the face image but also has an illumination-robust characteristic. By concatenating all the decomposed bit-planes, we generate an HD feature vector with an improved discriminative ability. To reduce the computational complexity while preserving the incorporated local structural information, a supervised dimension reduction method, the orthogonal linear discriminant analysis, is applied to the HD feature vector. Extensive experimental results show that existing classifiers with the proposed feature outperform those with other conventional features under various illumination, pose, and expression variations.     

Discriminative low-rank embedding with manifold constraint for image feature extraction and classification

The robustness against noise, outliers, and corruption is a crucial issue in image feature extraction. To address this concern, this paper proposes a discriminative low-rank embedding image feature extraction algorithm. Firstly, to enhance the discriminative power of the extracted features, a discriminative term is introduced using label information, obtaining global discriminative information and learning an optimal projection matrix for data dimensionality reduction. Secondly, manifold constraints are incorporated, unifying low-rank embedding and manifold constraints into a single framework to capture the geometric structure of local manifolds while considering both local and global information. Finally, test samples are projected into a lower-dimensional space for classification. Experimental results demonstrate that the proposed method achieves classification accuracies of 95.62%, 95.22%, 86.38%, and 86.54% on the ORL, CMUPIE, AR, and COIL20 datasets, respectively, outperforming dimensionality reduction-based image feature extraction algorithms.     

Pattern recognition using a family of design algorithms based uponthe generalized probabilistic descent method

This paper provides a comprehensive introduction to a novel approach to pattern recognition which is based on the generalized probabilistic descent method (GPD) and its related design algorithms. The paper contains a survey of recent recognizer design techniques, the formulation of GPD, the concept of minimum classification error learning that is closely related to the GPD formalization, a relational analysis between GPD and other important design methods, and various embodiments of GPD-based design, including segmental-GPD, minimum spotting error training, discriminative utterance verification, and discriminative feature extraction. GPD development has its origins in basic pattern recognition and Bayes decision theory. It represents a simple but careful re-investigation of the classical theory and successfully leads to an innovative framework. For clarity of presentation, detailed discussions about its embodiments are provided for examples of speech pattern recognition tasks that use a distance-based classifier. Experimental results in speech pattern recognition tasks clearly demonstrate the remarkable utility of the family of GPD-based design algorithms     

Combined trellis coding and DFE through Tomlinson precoding

The authors propose and evaluate a receiver architecture which combines the power of a decision feedback equalizer (DFE) with trellis coding, while allowing for minimal decoding delay in such a way that the total gain of the system is additive. The system is based on a structure that transposes the feedback filter of the DFE into the transmitter and, for high-order constellations, provides negligible increase in transmitter power. The first known hardware realization of a high bit rate digital subscriber line (HDSL) system that achieves the coding gain provided by a trellis code in addition to the equalization gain provided by the DFE is presented. A system whose complexity of implementation is comparable to that of a typical DFE and an independent Viterbi decoder is proposed     

基于内容的视频镜头的分类

论文在研究了视频关键帧选取和特征提取技术的基础上,提出了一种基于内容的视频镜头分类方法,并将其应用于动漫/真人的视频镜头的分类,以检验所提方法的性能。实验首先提取了视频的语义特征,接着使用互信息对特征的有效性进行分析,最后使用支持向量机作为分类器,对特征分析的结果进行验证。     

Network application profiling with traffic causality graphs

A network application profiling framework is proposed that is based on traffic causality graphs (TCGs), representing temporal and spatial causality of flows to identify application programs. The proposed framework consists of three modules: the feature vector space construction using discriminative patterns extracted from TCGs by a graph‐mining algorithm; a feature vector supervised learning procedure in the constructed vector space; and an application identification program using a similarity measure in the feature vector space. Accuracy of the proposed framework for application identification is evaluated, making use of ground truth packet traces from seven peer‐to‐peer (P2P) application programs. It is demonstrated that this framework achieves an overall 90.0% accuracy in application identification. Contributions are twofold: (1) using a graph‐mining algorithm, the proposed framework enables automatic extraction of discriminative patterns serving as identification features; 2) high accuracy in application identification is achieved, notably for P2P applications that are more difficult to identify because of their using random ports and potential communication encryption. Copyright © 2014 John Wiley & Sons, Ltd.     

Efficient decision feedback equalization for sparse wireless channels

A new efficient decision feedback equalizer (DFE) appropriate for channels with long and sparse impulse response (IR) is proposed. Such channels are encountered in many high-speed wireless communications applications. It is shown that, in cases of sparse channels, the feedforward and feedback (FB) filters of the DFE have a particular structure, which can be exploited to derive efficient implementations of the DFE, provided that the time delays of the channel IR multipath components are known. This latter task is accomplished by a novel technique, which estimates the time delays based on the form of the channel input-output cross-correlation sequence in the frequency domain. A distinct feature of the resulting DFE is that the involved FB filter consists of a reduced number of active taps. As a result, it exhibits considerable computational savings, faster convergence, and improved tracking capabilities as compared with the conventional DFE. Note that faster convergence implies that a shorter training sequence is required. Moreover, the new algorithm has a simple form and its steady-state performance is almost identical to that of the conventional DFE.     

Feature discovering for image classification via wavelet-like pattern decomposition

In this paper, we propose a feature discovering method incorporated with a wavelet-like pattern decomposition strategy to address the image classification problem. In each level, we design a discriminative feature discovering dictionary learning (DFDDL) model to exploit the representative visual samples from each class and further decompose the commonality and individuality visual patterns simultaneously. The representative samples reflect the discriminative visual cues per class, which are beneficial for the classification task. Furthermore, the commonality visual elements capture the communal visual patterns across all classes. Meanwhile, the class-specific discriminative information can be collected by the learned individuality visual elements. To further discover the more discriminative feature information from each class, we then integrate the DFDDL into a wavelet-like hierarchical architecture. Due to the designed hierarchical strategy, the discriminative power of feature representation can be promoted. In the experiment, the effectiveness of proposed method is verified on the challenging public datasets.     

Adaptive Bayesian decision feedback equalizer for dispersive mobileradio channels

The paper investigates adaptive equalization of time-dispersive mobile radio fading channels and develops a robust high performance Bayesian decision feedback equalizer (DFE). The characteristics and implementation aspects of this Bayesian DFE are analyzed, and its performance is compared with those of the conventional symbol or fractional spaced DFE and the maximum likelihood sequence estimator (MLSE). In terms of computational complexity, the adaptive Bayesian DFE is slightly more complex than the conventional DFE but is much simpler than the adaptive MLSE. In terms of error rate in symbol detection, the adaptive Bayesian DFE outperforms the conventional DFE dramatically. Moreover, for severely fading multipath channels, the adaptive MLSE exhibits significant degradation from the theoretical optimal performance and becomes inferior to the adaptive Bayesian DFE     

Decision feedback equalization for CDMA in indoor wirelesscommunications

Commercial interest in Code Division Multiple Access (CDMA) systems has risen dramatically in the last few years. It yields a potential increase in capacity over other access schemes, because it provides protection against interference, multipath, fading, and jamming. Recently, several interference cancellation schemes for CDMA have been proposed but they require information about all interfering active users or some channel parameters. The authors present an adaptive fractionally spaced decision feedback equalizer (DFE) for a CDMA system in an indoor wireless Rayleigh fading environment. This system only uses information about the desired user's spreading code and a training sequence. An analysis on the optimum performance of the DFE receiver shows the advantages of this system over others in terms of capacity improvements. A simulation of this system is also presented to study the convergence properties and implementation considerations of the DFE receiver. Effects on the performance because of sudden birth and death of users in the CDMA system and bit error rate performance of the DFE receiver is also presented     

Dual attention interactive fine-grained classification network based on data augmentation

The key to fine-grained image classification is to find discriminative regions. Most existing methods only use simple baseline networks or low-recognition attention modules to discover object differences, which will limit the model to finding discriminative regions hidden in images. This article proposes an effective method to solve this problem. The first is a novel layered training method, which uses a new training method to enhance the feature extraction ability of the baseline model. The second step focuses on key regions of the image based on improved long short-term memory (LSTM) and multi-head attention. In the third step, based on the feature map obtained by the dual attention network, spatial mapping is performed by a multi-layer perceptron (MLP). Then the element-by-element mutual multiplication calculation of the channel is performed to obtain a feature map with finer granularity. Finally, the CUB-200-2011, FGVC Aircraft, Stanford Cars, and MedMNIST v2 datasets achieved good performance.     

特征的支持度与其分类能力的关系研究

频繁模式挖掘在分类问题中得到了广泛的应用,大量的工作利用频繁模式挖掘对分类问题进行特征选择,但对于为什么频繁模式挖掘可以在分类问题中进行有效的特征选择则缺乏系统的研究.为了为频繁模式挖掘在分类问题中的特征选择应用提供理论基础,需要确立特征的支持度与特征分类能力之间的关系,本文以特征的信息增益作为分类能力的评价准则,讨论其与特征支持度之间的联系.首先证明了信息增益是特征支持度的上凸函数;然后,在二类问题和多类问题情况下,分别证明了具有低支持度或高支持度的特征具有有限的信息增益,即具有低支持度或高支持度的特征具有有限的分类能力.最后,通过仿真实验验证了支持度与信息增益之间的关系,为频繁模式挖掘在分类问题中的应用提供了理论基础.     

Two's complement computation sharing multiplier and its applications to high performance DFE

We present a novel computation sharing multiplier architecture for two's complement numbers that leads to high performance digital signal processing systems with low power consumption. The computation sharing multiplier targets the reduction of power consumption by removing redundant computations within system by computation reuse. Use of computation sharing multiplier leads to high-performance finite impulse response (FIR) filtering operation by reusing optimal precomputations. The proposed computation sharing multiplier can be applicable to adaptive and nonadaptive FIR filter implementation. A decision feedback equalizer (DFE) was implemented based on the computation sharing multiplier in a 0.25-/spl mu/ technology as an example of an adaptive filter. The performance and power consumption of the DFE using a computation sharing multiplier is compared with that of DFEs using a Wallace-tree and a Booth-encoded multiplier. The DFE implemented with the computation sharing multiplier shows improvement in performance over the DFE using a Wallace-tree multiplier, reducing the power consumption significantly.     

A decision-feedback equalizer with pattern-dependent feedback formagnetic recording channels

A new nonlinear equalizer for high-density magnetic recording channels is presented. It has a structure of the decision-feedback equalizer (DFE) with a nonlinear model at the feedback section and a dynamic threshold detector. The feedback nonlinear model is a sequence of look-up tables (LUTs) indexed by time, and each table is addressed by a transition pattern formed by one future and ν past transitions. We call this new nonlinear equalizer the pattern-dependent DFE (PDFE). The feedback nonlinear model cancels the trailing nonlinear intersymbol interference (ISI), and then the data decision is made by considering the precursor nonlinear ISI caused by one future symbol. We propose a tap optimization criterion SNR_d for the PDFE which in effect tries to maximize the output signal to noise ratio, and derive a closed-form solution for the tap values. We compare the detection performance of PDFE with that of the DFE and the RAM-DFE on experimental channels. The RAM-DFE is a DFE with one large LUT at its feedback section. The results show that the PDFE yields a significant performance improvement over the DFE and the RAM-DFE. Also the PDFE derived in this paper achieves a superior performance compared with the PDFE derived by the minimum mean-square-error criterion     

基于改进YOLOv4-tiny的轻量化室内人员目标检测算法

深度学习在室内人员检测领域应用广泛,但是传统的卷积神经网络复杂度大且需要高算力GPU的支持,很难实现在嵌入式设备上的部署。针对上述问题,该文提出一种基于改进YOLOv4-tiny的轻量化室内人员目标检测算法。首先,设计一种改进的Ghost卷积特征提取模块,有效减少了模型的复杂度;同时,该文通过采用带有通道混洗机制的深度可分离卷积进一步减少网络参数;其次,该文构建了一种多尺度空洞卷积模块以获得更多具有判别性的特征信息,并结合改进的空洞空间金字塔池化结构和具有位置信息的注意力机制进行有效的特征融合,在提升准确率的同时提高推理速度。在多个数据集和多种硬件平台上的实验表明,该文算法在精度、速度、模型参数和体积等方面优于原YOLOv4-tiny网络,更适合部署于资源有限的嵌入式设备。     

Kernel-based feature extraction under maximum margin criterion

In this paper, we study the problem of feature extraction for pattern classification applications. RELIEF is considered as one of the best-performed algorithms for assessing the quality of features for pattern classification. Its extension, local feature extraction (LFE), was proposed recently and was shown to outperform RELIEF. In this paper, we extend LFE to the nonlinear case, and develop a new algorithm called kernel LFE (KLFE). Compared with other feature extraction algorithms, KLFE enjoys nice properties such as low computational complexity, and high probability of identifying relevant features; this is because KLFE is a nonlinear wrapper feature extraction method and consists of solving a simple convex optimization problem. The experimental results have shown the superiority of KLFE over the existing algorithms.     

全数字16QAM解调器设计与实现

正交移幅键控技术作为一种高效的传输方式得到了广泛的应用,AGC环、定时环、载波环和正交判决反馈均衡器是16QAM解调器实现的关键环节,对它们的原理进行了详细论述,并从硬件的角度对各个环节的实现进行了说明,运用MATLAB软件对各个环节进行了定点仿真,使用Verilog语言完成整个解调器的软件设计。不仅对用于硬件实现的FPGA芯片做了介绍,而且通过判断解调后的眼图表明解调效果较好。