Filter Bank Multi Research Articles

Cascaded maximum median-margin discriminant projection with its application to face recognition

This paper introduces a novel method called cascaded maximum median-margin discriminant projection (CMMDP) for robust face recognition. Initially, we introduce the maximum median-margin discriminant projection (MMDP) method, which aims to maximize the distances between class medians and their inter-class neighbors, while simultaneously minimizing the distances between class medians and their intra-class samples. To extract features at multiple levels, we extend the MMDP method to its deep version, known as CMMDP. The CMMDP algorithm incorporates the MMDP technique to learn multistage filter banks specifically designed for facial images. Afterward, we utilize straightforward binary hashing techniques for efficient indexing, along with block histograms for effective feature pooling. The CMMDP model is simple and easy to train, while taking full account of the global and local structure of the samples by maximizing the distances between samples that are dissimilar and minimizing the distances between samples that are similar after projection. This results in a low-dimensional data representation with better discriminative performance. We evaluate the performance of CMMDP on face datasets including AR, ORL, ExtYaleB and CMU PIE databases. The results indicate that CMMDP achieves superior performance than most advanced methods.

Lattice factorization based causal symmetric paraunitary matrix extension and construction of symmetric orthogonal multiwavelets

In this paper, we propose a lattice factorization based symmetric paraunitary matrix extension method to design a causal symmetric paraunitary multifilter banks(PUMFBs) and construct compactly supported symmetric orthogonal multiwavelets by using the method.First, we transform a Laurent polynomial matrix consisting of polyphase components of a given symmetric orthogonal multifilter into a Laurent polynomial matrix with a simple structure, and then obtain a lattice factorization based symmetric paraunitary extension matrix by factorizing the transformed Laurent polynomial matrix and extending the constant matrix of the factorization. We also construct symmetric orthogonal multiwavelets by the proposed symmetric paraunitary matrix extension method. Finally, some examples are given to illustrate the proposed symmetric paraunitary matrix extension method.

SpPCANet: a simple deep learning-based feature extraction approach for 3D face recognition

Abstract A Sparse Principal Component Analysis Network (SpPCANet) based feature extraction is proposed here for 3D face recognition. The network consists of three basic components: (1) Multistage sparse principal component analysis filters, (2) Binary hashing, and (3) Block-wise histogram computation. Here, the sparse principal component analysis is used to learn multistage filter banks at the convolution stage, which is followed by binary hashing for indexing and block-wise histogram for pooling. Finally, a linear support vector machine (SVM) is used for classifying the features extracted by SpPCANet. The proposed network SpPCANet is a lightweight deep learning network. Three well-known 3D face databases, namely, Frav3D, Bosphorus3D, and Casia3D, are used for validating the proposed system. This proposed network has been extensively studied by varying different parameters, such as the number of filters at the convolution layer and the size of filters at the convolution layer and size of non-overlapping blocks at the pooling layer. Handling all types of variation of faces available in Frav3D, Bosphorus3D, and Casia3D databases, the system has acquired 96.93%, 98.54%, and 88.80% recognition rates, respectively.

Raspberry Pi-based Finger Vein Recognition System using PCANet

Finger Vein Recognition System (FVRS) is a biometric technology that identifies or verifies an individual identity based on unique vein patterns. Compared with other biometrics, it is more secure, anti-forgery and hygiene. Thus, it successfully utilized in many authentications nowadays. A feature extraction process is mandatory in maximizing recognition rate of the system. However, majority of feature extraction techniques utilized in FVRS only can provide low-level features, which return an unsatisfied result. A simple deep learning network is proposed in this paper, namely Principal Component Analysis Network (PCANet). PCANet composed of three basic data processing components: 1) PCA filter, 2) binary hashing 3) histograms. PCA is employed for learning multistage filter banks. Binary hashing and block histograms are the steps for indexing and pooling. The features extracted are represented in block-wise histograms. In this paper, PCANet is compared with PCA in term of performance and effectiveness. PCANet retrieved an accuracy of 98.5% under limited training samples, an increase of 21.3% than that of PCA. Factors which impact PCANet are studied to identify the limitations of PCANet. Classifier utilized in FVRS is kNN with Euclidean distance. OpenCV image processing library and C++ language are used in the development. The results returned prove that PCANet performance does not easily influenced by intra-class variability and limited training datasets. Meanwhile, the research noticed about the difficulty in recognition as age increased. The performance evaluation shows that the accuracy of FVRS is 92.67% with the utilization of PCANet. Concluded that PCANet is a promising technique in finger vein recognition.

Hybrid descriptors and Weighted PCA-EFMNet for Face Verification in the Wild

Unconstrained face recognition is a complex task because of the numerous intra-class changes caused by lighting, occlusion, facial expressions and poses changes. Such variations affect considerably the performance of facial recognition systems, particularly those based on 2D information. In this research work, an efficient feature extraction method, called Gabor Local Binarized Statistical Image Features, is introduced. A new deep learning network is also developed to extract features relying on data processing components: (1) Cascaded Weighted Principal Component Analysis (WPCA) with Enhanced Fisher Model (WPCA-EFM); (2) binary hashing; and (3) Block-wise Histograms. The suggested architecture (Weighted PCA-EFM) is employed to learn multistage filter banks. Simple block histogram and binary hashing are applied for indexing and pooling, which facilitate the design and learning of our architecture for Face Verification in the Wild. Classification is finally performed using distance measure Cosine and support vector machine. Our experiments are carried out in real-world dataset: Labeled Faces in the Wild. Experimental findings demonstrate that the developed technique achieved high accuracy of 95%.

Multiple description image compression based on multiwavelets

Multiple description coding (MDC) is one of the source coding techniques to alleviate the problems of packet loss in the network. The decoder estimates the lost signals from received ones, based on the certain statistical correlation between descriptions. However, this correlation also leads to compression redundancy at the same time. Therefore, how to make efficient use of the introduced correlation has great importance in practical MDC approaches. In this paper, we propose a multiple description image coding scenario based on balanced multiwavelets. Two simple and effective methods to reconstruct the original image from partial descriptions are suggested. Furthermore, optimization criterion corresponding to this multiwavelet based system is provided. According to this criterion, we can choose appropriate multifilter banks to satisfy different demands. Experimental results show that the optimized multifilter banks in a simulated transform coding environment perform very well.

Canonical correlation analysis networks for two-view image recognition

In recent years, deep learning has attracted an increasing amount of attention in machine learning and artificial intelligence areas. Currently, many deep learning network-related architectures such as deep neural networks (DNNs), convolutional neural network (CNN), wavelet scattering network (ScatNet) and principal component analysis network (PCANet) have been proposed. The most effective network is PCANet, which has achieved promising performance in image classification, such as for face, object and handwritten digit recognition. PCANet can only handle data that are represented by single-view features. In this paper, we present a canonical correlation analysis network (CCANet) to address image classification, in which images are represented by two-view features. The CCANet learns two-view multistage filter banks by a canonical correlation analysis (CCA) method and constructs a cascaded convolutional deep network. Then, we incorporate filters with binaryzation and block-wise histogram processes to form the final depth structure. In addition, we introduce a variation of CCANet—dubbed RandNet-2—in which the filter banks are randomly generated. Extensive experiments are conducted using the ETH-80, Yale-B, and USPS databases for object classification, face classification and handwritten digits classification, respectively. The experimental results demonstrate that the CCANet algorithm is more effective than PCANet, RandNet-1 and RandNet-2.

Spectral detection with multistage granularity bands using filter bank techniques for CR applications

In this paper, we propose a method to detect unused spectrum using multistage Cosine Modulated Filter Bank (CMFB) technique for Cognitive Radio (CR) applications. The precision of spectral detection is increased when the number of sub-bands in the filter bank is increased, this increases the computational complexity of the filter bank structure. We suggest a multistage filter bank structure to reduce the computational complexity. Spectrum detection is done effectively in filter banks by proper choice of prototype filter with specified sub-band attenuation. The sub-bands detected in coarser granularity bands are sensed with finer granularity bands for precise spectral detection depending on the predefined threshold and probability of false alarm. The remaining sub-bands are not sensed further which in turn reduces the computational complexity. Through simulations it is shown that spectrum detection with multistage granularity bands maintains bandwidth efficiency and reliability in spectrum detection even at low SNR.

A Low-Complexity Multistage Polyphase Filter Bank For Wireless Microphone Detection in CR

In this paper, we propose a low-complexity multistage polyphase filter bank for the detection and estimation of center frequency of wireless microphone (WM) in television channels for cognitive radios. The detection precision is directly related to the number of subbands (granularity) in the filter bank. This implies that the computational complexity becomes high if the number of stages and granularity of the filter bank structure are increased. The novelty of the proposed method is the estimation of center frequency with higher precision and reduced computational complexity using the centroid method. The merit of the centroid method is that the presence of WM can be detected along with estimation of center frequency in the first stage, if the WM lies partly in one subband and partly in the adjacent subband. This single-stage detection and estimation of WM significantly reduces computational and hardware complexities as well as latency. However, if the WM appears anywhere exclusively within a single subband, the detection process can be completed in the second stage without ambiguity. A mathematical expression for calculating the center frequency of WM from the subband energy (power) using centroid method is also derived and presented. The proposed scheme is analyzed and validated through extensive simulations for the detection of WM. The error in estimation of center frequency in most of the cases is less than 4 % for SNR variations from 0 to ź20 dB.

PCANet: A Simple Deep Learning Baseline for Image Classification?

In this paper, we propose a very simple deep learning network for image classification that is based on very basic data processing components: 1) cascaded principal component analysis (PCA); 2) binary hashing; and 3) blockwise histograms. In the proposed architecture, the PCA is employed to learn multistage filter banks. This is followed by simple binary hashing and block histograms for indexing and pooling. This architecture is thus called the PCA network (PCANet) and can be extremely easily and efficiently designed and learned. For comparison and to provide a better understanding, we also introduce and study two simple variations of PCANet: 1) RandNet and 2) LDANet. They share the same topology as PCANet, but their cascaded filters are either randomly selected or learned from linear discriminant analysis. We have extensively tested these basic networks on many benchmark visual data sets for different tasks, including Labeled Faces in the Wild (LFW) for face verification; the MultiPIE, Extended Yale B, AR, Facial Recognition Technology (FERET) data sets for face recognition; and MNIST for hand-written digit recognition. Surprisingly, for all tasks, such a seemingly naive PCANet model is on par with the state-of-the-art features either prefixed, highly hand-crafted, or carefully learned [by deep neural networks (DNNs)]. Even more surprisingly, the model sets new records for many classification tasks on the Extended Yale B, AR, and FERET data sets and on MNIST variations. Additional experiments on other public data sets also demonstrate the potential of PCANet to serve as a simple but highly competitive baseline for texture classification and object recognition.

Biorthogonal Multiwavelets with Sampling Property and Application in Image Compression

This paper discusses biorthogonal multiwavelets with sampling property. In such systems, vector-valued refinable functions act as the sinc function in the Shannon sampling theorem, and their corresponding matrix-valued masks possess a special structure. In particular, for the multiplicity $$r=2$$r=2, a biorthogonal multifilter bank can be reduced to two scalar-valued filters. Moreover, if the vector-valued scaling functions are interpolating, three different concepts: balancing order, approximation order and analysis-ready order, will be shown to be equivalent. Based on this result, we introduce the transferring armlet order for constructing biorthogonal balanced multiwavelets with sampling property. Also, some balanced biorthogonal multiwavelets will be obtained. Finally, application of biorthogonal interpolating multiwavelets in image compression is discussed. Experiments show that for the same length, the biorthogonal multifilter bank is superior to the orthogonal case. Moreover, certain biorthogonal interpolating multiwavelets are also better than the classical Daubechies wavelets.

Recognition of Facial Expression via Kernel PCA Network

In this work, a kernel principle component analysis network (KPCANet) is proposed for classification of the facial expression in unconstrained images, which comprises only the very basic data processing components: cascaded kernel principal component analysis (KPCA), binary hashing, and block-wise histograms. In the proposed model, KPCA is employed to learn multistage filter banks. It is followed by simple binary hashing and block histograms for indexing and pooling. For comparison and better understanding, We have tested these basic networks extensively on many benchmark visual datasets ( such as the JAFFE [13] database, the CMU AMP face expression database, a part of the Extended Cohn-Kanade (CK+) database), The results demonstrate the potential of the KPCANet serving as a simple but highly competitive baseline for facial expression recognition.

Balanced Biorthogonal Multiwavelet with Symmetric/Antisymmetric Filter Banks

A class of the balanced biorthogonal multiwavelets was constructed by defining a specific matrix filter structure, in which the multifilter banks of multiwavelets have had the desired symmetry. Here, the multifilter banks have possess symmetric/antisymmetric, which resembled filters of scalar wavelet and have in favor of application, notwithstanding the multiwavelets constructed in this paper have lost the linear phase, so they have formed a new type of multiwavelets undoubtedly.

Balanced Orthogonal Multiwavelets with Symmetric/Antisymmetric Filter Banks

A class of the balanced orthogonal multiwavelets was constructed by defining a specific matrix filter structure, in which the multifilter banks of multiwavelets have had the desired symmetry. The multifilter banks have possessed symmetric/antisymmetric, which resembled the filters of scalar wavelet and have in favor of application, notwithstanding the multiwavelets constructed in this paper lose the linear phase, they have formed a new type of multiwavelets undoubtedly.

BALANCED INTERPOLATORY MULTIWAVELETS WITH MULTIPLICITY r

Vector-valued refinable interpolatory functions with multiplicity r are discussed in this paper. This kind of refinable functions have a sampling property like Shannon's sampling theorem, and corresponding matrix-valued refinable masks possess special structure. In the context of multiwavelets, some properties of multifilter banks related will be present. Based on these properties, it will be shown that there are no symmetric (or antisymmetric) vector-valued refinable functions with interpolatory property. In the practical application, multiwavelets are always required to possess a certain degree of smoothness, which is related to three different concepts: balancing order, approximation order and analysis-ready order. In the general case, three notions are different. But if the scaling function is interpolatory, three concepts will be verified to equal to each other. Finally, a complete characterization of multifilter banks {H, G} will also be given and it will be used to construct some new balanced multiwavelets with interpolatory property for case r = 2, corresponding to which, multifilter banks have rational coefficients.

APPROXIMATION AND BALANCING ORDERS FOR TOTALLY INTERPOLATING BIORTHOGONAL MULTIWAVELET SYSTEMS

We consider totally interpolating biorthogonal multiwavelet systems with finite impulse response two-band multifilter banks, a study balancing order conditions of such systems. Based on FIR and interpolating properties, we show that approximation order condition is completely equivalent to balancing order condition. Consequently, a prefiltering can be avoided if a totally interpolating biorthogonal multiwavelet system satisfies suitable approximation order conditions. An example with approximation order 4 is provided to illustrate the result.

Harmonic Analysis and Multistage Filter Design for a Large Bleach Production Facility

Modern sodium chlorate or sodium hypochlorite-making processes (commonly known as “bleach”) take salt and water through an electrolytic process. The electrolytic process requires a significant amount of dc power. Bleach production plants, equipped with rectifiers, can generate potentially damaging harmonic currents if not mitigated or controlled. This paper presents a combination of harmonic mitigation methods applied at a large bleach production facility. The harmonic mitigation methods consisted of transformer phase shifting and multistage harmonic filter banks applied at 34.5 kV to satisfy IEEE Standard 519-1992 harmonic distortion limits at the point of common coupling. The authors faced several challenges with the harmonic filter design that included various operating conditions within the facility as well as numerous utility substation loading and capacitor combinations.

Balanced multiple description subband coding based on multifilter banks

Multiple description coding is one of the source coding techniques to alleviate the problems of packet loss in network. The encoder generates several different descriptions of the source and the decoder can produce a useful reconstruction with any received subset of these descriptions. Most of the earlier works concentrated on redundancy and the average distortion of the single-description reconstructions, which is insufficient to describe the problem. In this paper, a metric of the coding system, balance eccentric modulus, is defined to evaluate the difference in quality between the individual reconstructions. Since the discrete wavelet transform becomes trivial polyphase transform when setting the balance eccentric modulus at zero, we introduce a multiwavelet based scheme instead. Finally, a sufficient condition for balanced multiwavelet based multiple description coding is presented.

Balanced Multiwavelets With Interpolatory Property

Balanced multiwavelets with interpolatory property are discussed in this paper. This kind of multiwavelets can have a sampling property like Shannon's sampling theorem. It has been shown that the corresponding matrix-valued refinable mask has special structure, and an orthogonal multifilter bank {H(z),G(z)} can be reduced to a scalar valued conjugate quadrature filter (CQF) a(z) . But it does not mean that any scalar CQF can form a "good" multifilter bank which can generate a vector-valued refinable function with some degree of smoothness. In the context of balanced multiwavelets, we give the definition of transferring balance order, which a scalar CQF a(z) satisfies, to guarantee that the multiwavelet Ψ generated is balanced. On the basis of the parametrization of a scalar CQF with any length and conditions of transferring balance order, parametrization of multifilter banks which can generate interpolatory multiwavelet and interpolatory scaling function, is gotten. Moreover, some balanced interpolatory multiwavelets have been constructed. Interpolatory analysis-ready multiwavelets (armlets) are also discussed in this paper. It is known that conditions of armlets are easy to validate, compared with balanced multiwavelets. But it will be present that if the corresponding scaling function Φ is interpolatory, the multiwavelet Ψ is balanced of order n if and only if it is an armlet of order n. Finally, the application of balanced multiwavelets with interpolatory property in image processing is also discussed.

Balanced Multifilter Banks for Multiple Description Coding

The parametrization for one kind of multifilter banks generating balanced multiwavelets is presented in this paper, in which two lowpass filters are flipping filters, and two highpass filters have linear phase. Based on these parametric expressions, some balanced multiwavelets and analysis-ready multiwavelets are constructed, which are symmetric, or antisymmetric. Moreover, on the basis of balanced multiwavelet transform, a new method of multiple description coding is given, and experiments show that this method works well. Compared with the traditional multiple description coding method, this method has low redundancy.