Minimum Reconstruction Error Research Articles

On-Board Signal Quality Assessment Guided Compression of Photoplethysmogram for Personal Health Monitoring

Photoplethysmography (PPG) is a popular diagnostic tool for the assessment of various cardiovascular functions. Under continuous ambulatory measurements, PPG data get corrupted due to motion artifact (MA). Thus, on-board pulse signal quality assessment (SQA) before transmission can save the battery life of the wearable device for portable health monitoring applications. This article describes an SQA guided compression (SQAGC) of PPG data using a modified gain-shaped vector quantization (GSVQ) technique. The SQA was performed using kurtosis and autocorrelation to generate a binary classification rule to detect good quality pulses. Only these were considered for further compression. A notable contribution is reconstruction error minimization using the extracted features from the residual signal using a deep autoencoder (DAE), achieving a low percentage root-mean-squared difference (PRD). The SQAGC technique was evaluated using public databases like MIMIC-II, BIDMC, and PRRB as well as with real volunteers’ PPG collected in the laboratory environment. The SQA achieved an accuracy of 96.5% to identify good quality PPG segments out of expert annotated 9200 beats. The compression factor (and PRD) with 400 min duration data from Physionet MIMIC-II, BIDMC, PRRB, and volunteers’ data were 15.8 (and 0.31), 15.7 (and 0.21), 17.8 (and 0.33), and 18.2 (and 0.59), respectively, using 12-bit resolution and 125 Hz sampling. A real-time on-device implementation using quad-core ARM Cortex-A53, 1.2 GHz, supported by 1 GB RAM, achieved a latency of 546 ms with 327 kB of memory engagement for a 3 s PPG window. The compression ratio (CR) achieved comparable results, while PRD outperforms the published results using MIMIC-II data set.

PRIMAL-GMM: PaRametrIc MAnifold Learning of Gaussian Mixture Models.

We propose a ParametRIc MAnifold Learning (PRIMAL) algorithm for Gaussian mixtures models (GMM), assuming that GMMs lie on or near to a manifold of probability distributions that is generated from a low-dimensional hierarchical latent space through parametric mappings. Inspired by principal component analysis (PCA), the generative processes for priors, means and covariance matrices are modeled by their respective latent space and parametric mapping. Then, the dependencies between latent spaces are captured by a hierarchical latent space by a linear or kernelized mapping. The function parameters and hierarchical latent space are learned by minimizing the reconstruction error between ground-truth GMMs and manifold-generated GMMs, measured by Kullback-Leibler Divergence (KLD). Variational approximation is employed to handle the intractable KLD between GMMs and a variational EM algorithm is derived to optimize the objective function. Experiments on synthetic data, flow cytometry analysis, eye-fixation analysis and topic models show that PRIMAL learns a continuous and interpretable manifold of GMM distributions and achieves a minimum reconstruction error.

DEVELOPING A COMPUTER MODEL OF WAVELET FILTERING AND RECONSTRUCTION OF ONE-DIMENSIONAL SIGNALS

In recent decades, wavelet functions have been used in radio engineering and communication systems for processing complex signals and images in parallel with the Fourier transform methods. Based on orthogonal bases, wavelets occupy an intermediate position between a sinusoid and an impulse function and, due to their unique properties, can approximate any function. Algorithms for analysis and reconstruction, as well as wavelet filtering of radio engineering signals, are presented, which use the best wavelet bases, optimal decomposition levels, effective thresholding methods and optimal threshold values selected as a result of research. Programs have been developed which implement the above algorithms, which calculate the estimating parameters of the analyzed signals, which make it possible to evaluate the effectiveness of the algorithms. Based on the research results, a computer model has been developed which performs analysis and reconstruction, as well as wavelet filtering of a one-dimensional signal, providing minimal reconstruction errors and effective noise removal.

Sparse PCA via l2,p-Norm Regularization for Unsupervised Feature Selection.

In the field of data mining, how to deal with high-dimensional data is an inevitable topic. Since it does not rely on labels, unsupervised feature selection has attracted a lot of attention. The performance of spectral-based unsupervised methods depends on the quality of the constructed similarity matrix, which is used to depict the intrinsic structure of data. However, real-world data often contain plenty of noise features, making the similarity matrix constructed by original data cannot be completely reliable. Worse still, the size of a similarity matrix expands rapidly as the number of samples rises, making the computational cost increase significantly. To solve this problem, a simple and efficient unsupervised model is proposed to perform feature selection. We formulate PCA as a reconstruction error minimization problem, and incorporate a l2,p-norm regularization term to make the projection matrix sparse. The learned row-sparse and orthogonal projection matrix is used to select discriminative features. Then, we present an efficient optimization algorithm to solve the proposed unsupervised model, and analyse the convergence and computational complexity of the algorithm theoretically. Finally, experiments on both synthetic and real-world data sets demonstrate the effectiveness of our proposed method.

Compressive-sensing based damped multi-mode extraction with nonuniform spatial sampling

Multi-mode extraction is an important technique providing essential information for waveguide model validation and waveguide condition assessment. In this paper, an extraction method for damped multi-modes is proposed using randomly distributed nonuniform spatial samples. It addresses the problems of the conventional measurement method using uniform sampling, such as high sampling rate and large sample numbers. The proposed method is based on enhanced matrix completion (EMaC) which constructs a Hankel matrix by spatial signal. Using the measured nonuniform spatial samples, EMaC uses nuclear norm compressive sensing (CS) to reconstruct all entries of the Hankel matrix, from which the spatial signal is recovered and modes with their wavenumbers and attenuations are extracted. The bounds in terms of minimum sample numbers, spectrum separation and reconstruction error are analyzed by theory and quantified by simulations. Using the simulated data of a thin Titanium plate waveguide and the experimental data of a water-filled steel pipeline, the proposed method is shown to extract damped multi-modes using just 25% of the whole uniform samples set. The accuracy of wavenumber and attenuation estimates of the proposed method are found to be similar to the conventional method using uniform samples.

Class factorized complex variational auto-encoder for HRR radar target recognition

In the field of radar automatic target recognition (RATR), the high-resolution range profile (HRRP) has received intensive attention. Bar a few exceptions, almost all HRRP-based ATR classification systems ignore the phase of the HRRP when the data is input to the classifier, relying instead only on the magnitude of the complex HRRP samples. This approach ignores the phase of the complex HRRPs, which reduces the information in the signal. In this paper, we develop a novel class factorized complex variational auto-encoder (CFCVAE) to utilize the phase of the high range resolution (HRR) radar echo for recognition. The CFCVAE is a complex-valued neural network (CV-NN) consisting of the encoding and decoding modules. In CFCVAE, the encoding module projects the observed data into the latent space, and then the latent features are fed to the decoding module, which further maps the latent features to data. In particular, the decoding module introduces the class labels to partition the whole observations into some parts, each of which is depicted by a specific class-decoder. Compared with the traditional variational auto-encoder (VAE) containing a single decoder, the CFCVAE can give a more accurate description to the whole dataset via multiple class-decoders, thus improving the characterization ability of features. In addition, based on the class labels, the CFCVAE employs the conditional prior on the latent variable to enhance the discrimination of features. Moreover, a complex backpropagation algorithm is derived for CFCVAE training, and a sample is classified to the class corresponding to the class-decoder with the minimum reconstruction error in the test stage. Experimental evaluations on the measured data indicate that the proposed method indeed achieves very promising target recognition performance.

Low-rank adaptive graph embedding for unsupervised feature extraction

Most of manifold learning based feature extraction methods are two-step methods, which first construct a weighted neighborhood graph and then use the pre-constructed graph to perform subspace learning. As a result, these methods fail to use the underlying correlation structure of data to learn an adaptive graph to preciously characterize the similarity relationship between samples. To address this problem, we propose a novel unsupervised feature extraction method called low-rank adaptive graph embedding (LRAGE), which can perform subspace learning and adaptive probabilistic neighborhood graph embedding simultaneously based on reconstruction error minimization. The proposed LRAGE is imposed with low-rank constraint for the sake of exploring the underlying correlation structure of data and learning more informative projection. Moreover, the L2,1-norm penalty is imposed on the regularization to further enhance the robustness of LRAGE. Since the resulting objective function has no closed-form solutions, an iterative optimization algorithm is elaborately designed. The convergence of the proposed algorithm is proved and the corresponding computational complexity analysis is also presented. In addition, we explore the potential properties of the proposed LRAGE by comparing it with several similar models on both synthetic and real-world data sets. Extensive experiments on five well-known face data sets and three non-face data sets demonstrate the superiority of the proposed LRAGE.

Denoising of Hydrogen Evolution Acoustic Emission Signal Based on Non-Decimated Stationary Wavelet Transform

Monitoring the evolution of hydrogen gas on carbon steel pipe using acoustic emission (AE) signal can be a part of a reliable technique in the modern structural health-monitoring (SHM) field. However, the extracted AE signal is always mixed up with random extraneous noise depending on the nature of the service structure and experimental environment. The noisy AE signals often mislead the obtaining of the desired features from the signals for SHM and degrade the performance of the monitoring system. Therefore, there is a need for the signal denoising method to improve the quality of the extracted AE signals without degrading the original properties of the signals before using them for any knowledge discovery. This article proposes a non-decimated stationary wavelet transform (ND-SWT) method based on the variable soft threshold function for denoising hydrogen evolution AE signals. The proposed method filters various types of noises from the acquired AE signal and removes them efficiently without degrading the original properties. The hydrogen evolution experiments on carbon steel pipelines are carried out for AE data acquisition. Simulations on experimentally acquired AE signals and randomly generated synthetic signals with different levels of noise are performed by the ND-SWT method for noise removal. Results show that our proposed method can effectively eliminate Gaussian white noise as well as noise from the vibration and frictional activity and provide efficient noise removal solutions for SHM applications with minimum reconstruction error, to extract meaningful AE signals from the large-scale noisy AE signals during monitoring and inspection.

Sparse Representation Classification With Structured Dictionary Design Strategy for Rotating Machinery Fault Diagnosis

Fault diagnosis technique is the core of Prognostics and Health Management (PHM) system, which plays a crucial role in the intelligent operation and maintenance of various rotating machineries. In this paper, we present a novel sparse representation classification framework with structured dictionary design strategy (SRC-SDD) for intelligent fault diagnosis of rotating machineries. The proposed SRC-SDD method consists of two stages, i.e., the structured dictionary design stage and the sparsity-based intelligent diagnosis stage. In the first stage, the novelty of SRC-SDD lies in the overlapping segmentation strategy for structured dictionary design, which leverages the structured prior knowledge of rotating machinery vibration signals, namely, the periodic self-similarity and shift-invariance properties. In the second stage, SRC-SDD achieves fault recognitions of testing samples using a sparsity-based diagnosis strategy based on the minimum sparse reconstruction error. The proposed structured dictionary design strategy can enhance the representation power of dictionaries and thus promote the recognition performance of the sparsity-based diagnosis strategy. Finally, the effectiveness of SRC-SDD has been validated on the gearbox fault dataset from IEEE PHM society. The diagnosis results show that SRC-SDD achieves the excellent recognition accuracy of 100% for predicting six different gearbox health states. Further, the comparative studies with three conventional SRC methods prove the superiority of SRC-SDD in terms of both the recognition performance and computation efficiency.

Dual robust regression for pattern classification

Linear regression-based and extended methods have been widely used in pattern classification. These methods can be roughly divided into two categories: reconstruction error-based methods and discriminative methods. The reconstruction error-based methods search the target label by learning the representation coefficients to compute the minimal reconstruction error. The goal of the discriminative methods is to learn the projection matrix to predict the target label of the image. To combine the advantages of these two kinds of regression-based methods, this paper presents a dual robust regression framework (DualRR) for pattern classification. In the training stage, a double low-rank robust regression model (DLR) is proposed to learn the projection matrix. In DLR, low-rank robust regression motivates us to model the data as the sum of a low-rank clean data and sparse noise matrix. The low-rank is further used to constrain the projection matrix to enhance the discriminative performance. In the testing stage, the proposed framework employs a robust regression representation model to learn the optimal representation coefficients and obtain the reconstruction sample to approximate the test sample. We thus apply the reconstruction sample to search the classification label by using the projection matrix learned in the training stage. Extensive experiments are conducted on six public available databases, namely, LFW, FRGC, CUHK Sketch, PolyU Palm, NUST-RF and Caltech 101, demonstrating the merits of the proposed model over state-of-the-art regression-based classification methods.

Target redirected regression with dynamic neighborhood structure

Least squares regression (LSR) has attracted widespread attention in the fields of statistics, machine learning, and pattern recognition. However, it utilizes strict zero-one regression targets, which leads to inferior performance on classification tasks. Furthermore, LSR ignores the local manifold structures of data and lacks robustness. To address these issues, this paper proposes a general regression framework called RLRR, where a low-rank constraint is imposed on regression matrices to explore the underlying correlation structures of classes. Strict zero-one regression targets are redirected to more feasible variable matrices for the purpose of margin amplification of different classes. Additionally, rather than using a pre-constructed weighted graph, the proposed framework dynamically updates the neighborhood structures of data to preserve original manifold structures. By utilizing this framework as a general platform, we developed two dynamic neighborhood-structure-based regression models called RLRRM and RLRRP. RLRRM integrates a reconstruction error minimization term into the proposed RLRR framework, whereas RLRRP aims to preserve the local geometric structures of data in a low-dimensional subspace. Both RLRRM and RLRRP use theℓ2,1-norm penalty to replace the traditional F-norm penalty for the projection matrix for the sake of self-adaptive feature selection. Instead of directly solving the resultant optimization problems with non-convex constraints, we adopt the variable-splitting and penalty techniques to derive an equivalent solution. Analysis of the corresponding convergence and computational complexity characteristics is also presented. Extensive experiments on several well-known datasets demonstrate the promising performance of the proposed models.

The Contrast-Enhanced Ultrasound Information Technology in the Treatment of Liver Tumors

Objective: To study the automatic diagnosis method of liver tumors in the contrast-enhanced ultrasound environment, assist doctors in the clinical diagnosis of liver tumors intuitively, conveniently, and accurately, thereby improve the cure rate of liver tumors. Methods: First, six sets of experimental data were constructed. The automatic diagnosis experiment of liver tumors through contrast-enhanced ultrasound was performed by the combination of sparse representation-based support vector machine (SVM) and principal component analysis (PCA)-based SVM, as well as the sparse representation classification method. The effect of classification decision principles on experiments was further studied. Results: The SVM method had an average effect on diagnostic accuracy. The average diagnostic accuracy of the six different experimental data sets was 76%, and the average diagnosis time was 300 s. The feature extraction method based on the combination of sparse representation and PCA was applied to the SVM method to achieve an optimal diagnosis. The average diagnosis accuracy rate could reach 87%, and the average time was more than 1,000 s. Using the sparse classification representation method, the diagnostic accuracy rate for the six experimental data sets constructed was above 93%, with a maximum of 99%, and the average time was 210 s. The sparse classification representation using the principle of minimum reconstruction error classification decision had an average diagnostic accuracy rate of 99% and an average time of 128 s. Conclusion: The sparse classification representation for the clinical diagnosis of liver tumors by contrast-enhanced ultrasound had high accuracy and consumed less time. Therefore, the constructed method was valuable.

Approach to model human appearance based on sparse representation for human tracking in surveillance

In human tracking, sparse representation successfully localises the human in a video with minimal reconstruction error using target templates. However, the state-of-the-art approaches use colour and local appearance of a human to discriminate the human from the background regions, and hence fail when the human is occluded and appears in the varying illumination environment. In this study, a robust tracking algorithm is proposed that utilises gradient orientation and fine and coarse sparse representation of the target template. Sparse representation-based human appearance model utilises weighted gradient orientation that is insensitive to illumination variation. Coarse and fine representation of sparse code facilitates tracking under varying scales. Subspace learning from image gradient orientation is enforced with occlusion detection during the dictionary updation stage to capture the visual characteristics of the local human appearance that supports tracking under partial occlusion with lesser tracking error. The proposed human tracking algorithm is evaluated on various datasets and shows efficient human tracking performance when compared to the other state-of-the-art approaches. Furthermore, the proposed human tracking algorithm is suitable for surveillance applications.

Compressed encoding of vibration signals using extremum sampling

In this paper, we propose extremum sampling, an activity-dependent scheme for sampling and encoding of vibration signals in a compressed form. Vibration signals are generally utilized for the condition monitoring of mechanical systems, including launch vehicles. With the highest frequency among all the other sensor outputs, enormous data is collected if vibration signals are sampled at the Nyquist rate, leading to complicated analysis. Besides, there is a limitation of the available bandwidth in launch vehicles and wireless nodes. Conventional data compression methods have to operate on all the samples acquired at the uniform rate, which increases the computational complexity at the transmitter. The proposed extremum sampling method is based on level-crossing sampling, and hence it has the inherent noise tolerance depending upon the value of the inter-level distance. A sample is taken at a level-crossing instant, if there is a change of direction of the signal compared to the direction at the previous level-crossing instant. Compared to compressed-sensing, extremum sampling offers a simple method for the real-time acquisition and reconstruction of the signal. Excellent compression ratios have been achieved with minimal reconstruction errors. Since the method accomplishes compression during the sampling process itself, the additional burden at the transmitter for subjecting all the acquired samples to a compression algorithm is evaded. The generation of fewer samples results in the reduction of dynamic power consumption at the transmitter, memory, and bandwidth requirements.

An Efficient Compressive Sensing Routing Scheme for Internet of Things Based Wireless Sensor Networks

Internet of Things (IoT) integrates diverse types of sensors, mobiles and other technologies to physical world and IoT technology is used in a wide range of applications. Compressive sensing based in-network compression is an efficient technique to reduce communication cost and accurately recover sensory data at the base station. In this paper, we investigate how compressive sensing can be combined with routing protocols for energy efficient data gathering in IoT-based wireless sensor networks. We propose a new compressive sensing routing scheme that includes the following new algorithms: (1) seed estimation algorithm to find the best measurement matrix by selecting the best-estimated seed, (2) chain construction algorithm to organize the network nodes during transmitting and receiving process, (3) compression approach to reduce the energy consumption and prolong the network lifetime by reducing the local data traffic, and (4) reconstruction algorithm to reconstruct the original data with minimum reconstruction error. The simulation results reveal that the proposed scheme outperforms existing baseline algorithms in terms of energy consumption, network lifetime and reconstruction error.

Deep subspace clustering to achieve jointly latent feature extraction and discriminative learning

Deep subspace clustering based on autoencoder and self-expression layer has become popular for clustering method. However, these models only focus on feature extraction by minimizing reconstruction error rather than specific clustering task, leading to unsatisfactory performance. To overcome the above shortcoming, we propose a deep subspace clustering framework which jointly extracts features via the embedding neural network and performs subspace learning. The model contains three modules, an autoencoder, a self-expression layer and a supervised competitive feature learning module. The proposed model is highly capable of capturing characteristic features to guide the clustering task by using relative entropy to minimize probabilistic cluster assignments and the target variables. The three modules are consolidated to be jointly trained and optimized competitively. Experimental results on five benchmark datasets demonstrate the effectiveness of the proposed deep subspace clustering by comparing with eleven baseline methods.

A new L1-LRC based model for oranges origin identification with near infrared spectra data

In order to establish an accurate and efficient model for geographical origin identification of oranges, a new model based on L1-norm linear regression classification (L1-LRC) is proposed. The proposed L1-LRC for orange origin identification is based on minimum reconstruction error using the L1-norm regularization learning method, which can combine the feature selection and classifier learning, and can reveal the structure characteristics of spectral information effectively. The experimental results show that the proposed L1-LRC model can achieve higher accuracy rate of 92.35% and perform much better than existing models when using only a few training samples. Thus, this work would lead to a new method for fast and efficient identification of geographical origins with near infrared (NIR) spectroscopy.

Tracking objects with partial occlusion by background alignment

Visual object tracking is a challenging and fundamental research topic in the field of computer vision. Recent years, many subspace-learning based methods have been proposed for visual object tracking with promising results. These methods reconstruct candidate states by a set of basis vectors and select the best state with minimum reconstruction error. It is well known that the accuracy of reconstructed results is seriously affected by partial occlusion. Besides, updating with occlusions is likely to mislead the tracker to drift away. Existing methods either do not consider these situations or find occlusion regions only by current observation and reconstruction. In fact, occlusion regions usually come from the background regions in previous frames, which are neglected in existing methods. Under this assumption, a novel object tracking algorithm called Partial Occlusion by Background Alignment (POBA) is proposed, which aims to find the best candidate state with an accurate occlusion mask. The POBA tracker treats current observation as a combination of object appearance and occlusion regions. The object appearance is modelled by basis vectors through incremental PCA over grayscale images. Then the occlusion region is reconstructed from last frame under the assumption that the backgrounds between two consecutive frames are almost identical. Besides, most candidate states are different from the object obviously, which can be filtered by some predefined occlusion masks so that computational complexity can be further reduced. Finally, the POBA tracker was analyzed on 8 challenge sequences and evaluated on two challenging datasets including OTB2015 and Temple Color. It gets an AUC of 0.456, a success rate score of 0.538 and a precision score of 0.626 in OPE on OTB2015 dataset. All indicators are increased by more than 23% compared with the 6 classical trackers.

Robust neighborhood embedding for unsupervised feature selection

Unsupervised feature selection is an efficient approach of dimensionality reduction for alleviating the curse of dimensionality in the countless unlabeled high-dimensional data. In view of the sparseness of the high-dimensional data, we propose a robust neighborhood embedding (RNE) method for unsupervised feature selection. First, with the fact that each data point and its neighbors are close to a locally linear patch of some underlying manifold, we obtain the feature weight matrix through the locally linear embedding (LLE) algorithm. Second, we use ℓ1-norm to describe reconstruction error minimization, i.e., loss function to suppress the impact of outlier and noises in the dataset. As the RNE model is convex but non-smooth, we exploit alternation direction method of multipliers (ADMM) to solve it. Finally, extensive experimental results on benchmark datasets validate that the RNE method is effective and superior to the state-of-the-art unsupervised feature selection algorithms in terms of clustering performance.

Human Action Recognition Based on Supervised Class-specific Dictionary Learning with Deep Convolutional Neural Network Features

Human action recognition under complex environment is a challenging work. Recently, sparse representation has achieved excellent results of dealing with human action recognition problem under different conditions. The main idea of sparse representation classification is to construct a general classification scheme where the training samples of each class can be considered as the dictionary to express the query class, and the minimal reconstruction error indicates its corresponding class. However, how to learn a discriminative dictionary is still a difficult work. In this work, we make two contributions. First, we build a new and robust human action recognition framework by combining one modified sparse classification model and deep convolutional neural network (CNN) features. Secondly, we construct a novel classification model which consists of the representation-constrained term and the coefficients incoherence term. Experimental results on benchmark datasets show that our modified model can obtain competitive results in comparison to other state-of-the-art models.