Blind Segmentation Research Articles

A high‐speed unsupervised hardware architecture for rapid diagnosis of COVID‐19

SummaryIn the diagnosis of COVID‐19, investigation, analysis, and automatic counting of blood cell clusters are the most essential steps. Currently employed methods for cell segmentation, identification, and counting are time‐consuming and sometimes performed manually from sampled blood smears, which is hard and needs the support of an expert laboratory technician. The conventional method for the blood‐count‐test is by automatic hematology analyzer which is quite expensive and slow. Moreover, most of the unsupervised learning techniques currently available presume the medical practitioner to have a prior knowledge regarding the number and action of possible segments within the image before applying recognition. This assumption fails most often as the severity of the disease gets increased like the advanced stages of COVID‐19, lung cancer etc. In this manuscript, a simplified automatic histopathological image analysis technique and its hardware architecture suited for blind segmentation, cell counting, and retrieving the cell parameters like radii, area, and perimeter has been identified not only to speed up but also to ease the process of diagnosis as well as prognosis of COVID‐19. This is achieved by combining three algorithms: the K‐means algorithm, a novel statistical analysis technique‐HIST (histogram separation technique), and an islanding method an improved version of CCA algorithm/blob detection technique. The proposed method is applied to 15 chronic respiratory disease cases of COVID‐19 taken from high profile hospital databases. The output in terms of quantitative parameters like PSNR, SSIM, and qualitative analysis clearly reveals the usefulness of this technique in quick cytological evaluation. The proposed high‐speed and low‐cost architecture gives promising results in terms of performance of 190 MHz clock frequency, which is two times faster than its software implementation.

Blind sidewalk segmentation based on the lightweight semantic segmentation network

A lightweight semantic segmentation network is proposed to solve the problem of real-time segmentation of blind sidewalk. On the basis of U-Net’s encode-decoding structure, the inverse residual block composed of deep separable convolution is used to replace the ordinary convolution for feature extraction, and the number of convolution layer channels is compressed. Experiments show that the blind segmentation method used can effectively overcome the disadvantages of traditional methods that are affected by the environment easily. Compared with U-Net, the parameter amount is reduced by 25.2 times, and the reasoning speed is increased by 3.6 times, while the loss of precision is just less than 2%, which basically meet the requirement of real-time inference.

Radon spectrogram-based approach for automatic IFs separation

The separation of overlapping components is a well-known and difficult problem in multicomponent signals analysis and it is shared by applications dealing with radar, biosonar, seismic, and audio signals. In order to estimate the instantaneous frequencies of a multicomponent signal, it is necessary to disentangle signal modes in a proper domain. Unfortunately, if signal modes supports overlap both in time and frequency, separation is only possible through a parametric approach whenever the signal class is a priori fixed. In this work, time-frequency analysis and Radon transform are jointly used for the unsupervised separation of modes of a generic frequency modulated signal in noisy environment. The proposed method takes advantage of the ability of the Radon transform of a proper time-frequency distribution in separating overlapping modes. It consists of a blind segmentation of signal components in Radon domain by means of a near-to-optimal threshold operation. The inversion of the Radon transform on each detected region allows us to isolate the instantaneous frequency curves of each single mode in the time-frequency domain. Experimental results performed on constant amplitudes chirp signals confirm the effectiveness of the proposed method, opening the way for its extension to more complex frequency modulated signals.

Phase-domain Deep Patient-ECG Image Learning for Zero-effort Smart Health Security.

Smart health is quickly boosted by technological advancements: smart sensors, body sensor network, internet of medical things and big data. Vast amounts of smart health big data from ubiquitous sensors pose unprecedented challenges to the security and privacy protection, which is extremely critical in healthcare applications. The vital signs, user daily behaviors, medicine recommendations, and so many other health data are vulnerable to different attacks, due to the fact that wearable/mobile monitors have very strict performance/power constraints, which limit the complexity of security protocols. In this paper, we study how to leverage a natural vital signal (Electrocardiogram - ECG) for user identification purpose, without introducing new hardware sensing devices. ECG is not only a gold standard cardiac signal, but also unique to each individual. We investigate a phase-domain deep patient-ECG image learning framework, to tackle key challenges in ECG biometric user identification: high diversities of ECG morphologies due to heart diseases, and time-consuming/ineffective heartbeat localization methods & manual feature engineering. The ultimate goal is to make the smart health security zero-effort: use `phase-domain transformation' to enable blind signal segmentation without localizing heartbeats; create a computer image processing-like task by `pixelating' phase-domain ECG trajectories to ECG images; and enable automatic (non-manual) `deep feature learning' using a deep convolutional neural network. Evaluated on two patient-ECG databases, this zero-effort framework achieves an accuracy as high as 97.2%, and greatly outperforms state-of-the-art studies in terms of the generalization ability and/or performance. This study is expected to enable highly challenging patient-ECG biometric user identification, by generalizable blind signal segmentation and deep feature learning strategies, in the era of smart health boosted by internet of medical things and big medical data.

A New Deep Segmentation Quality Assessment Network for Refining Bounding Box Based Segmentation

Capturing quality cues in blind segmentation quality assessment (SQA) is a challenging task. This paper proposes a new blind SQA method that captures a variety of quality cues using two considerations: one is to consider different types of data, such as segmentation mask and original image patch and the other is to consider different types of data operations, such as the max pooling operation, the average pooling operation, and the convolution operation. An end-to-end segmentation quality assessment network is proposed to capture these different types of quality cues by the above-mentioned two considerations. Then, since the segmentation result by the traditional bounding box-based segmentation method using fixed parameter setting is usually inaccurate, we use the proposed blind SQA network to refine the traditional bounding box-based segmentation results, with the idea of selecting parameters that are adaptive to each image. A two-step refinement method is proposed. The first step generates multiple segmentation results by various parameter settings. The second step, then, uses the segmentation quality assessment to select the best quality segmentation result as the new result. To train and verify the proposed method, a new dataset constructed by using the PASCAL VOC dataset and the bounding box-based segmentation methods, such as Grabcut and improved FCN method, is constructed. The proposed method is finally verified on the constructed dataset and two tasks: the segmentation quality assessment and the bounding box-based segmentation. The experimental results demonstrate the effectiveness of the proposed method.

ECG authentication system design incorporating a convolutional neural network and generalized S-Transformation

Electrocardiogram (ECG) is gaining increased attention as a biometric method in a wide range of applications, such as access control and security/privacy requirements. The majority of reported investigations using the ECG biometric method are usually based on fiducial or nonfiducial methods, which are always accompanied by a series of issues, such as locating fiducial points accurately is difficult, feature selection is subjective, and classifiers are limited by the quantity and structure of data. This paper proposes a new biometric authentication system for human identification that uses ECG signals as a biometric trait and integrates a generalized S-transformation and a convolutional neural network (CNN). Specifically, we first introduce a blind segmentation strategy that effectively avoids difficult data-specific heartbeat recognition and segmentation techniques. Then, a generalized S-transformation is performed on the blind signal-processed ECG signal, capturing the ECG trajectory at each time point in the frequency domain. Next, the getframe technology is used to capture an image of the ECG trajectories and convert the one-dimensional signal to a two-dimensional image, which serves as the input layer of the CNN, thus fully reflecting the changing trend in the ECG signal spectrum characteristics over a continuous period. Finally, the CNN is used for automatic discriminative feature learning and representations, which avoids a tedious feature extraction algorithm. In addition, considering the possible impact of ECG signals with different signal behaviors on identification, experiments are performed on three ECG databases with diverse features, comprising normal individuals, atrial fibrillation patients, and a noisy database, to evaluate the effectiveness of the proposed algorithm. Promising identification rates of 99%, 98%, and 99% were achieved, respectively. Thus, our proposed ECG authentication system can be effectively used for identity recognition under various conditions.

Automatic Fuzzy License Plate Recognition Based on Deep Learning

At present, the clear license plate recognition algorithm has been mature, but for the blurred license plate which cannot be recognized by the human eye, the recognition rate of the traditional license plate recognition algorithm is too low. In view of this, a license plate character recognition algorithm based on convolutional neural network structure is proposed. In this study, a large number of training samples of fuzzy characters are constructed, and a convolutional neural network is trained for blind segmentation of fuzzy license plate characters. By calling the trained convolutional neural network, we can recognize the characters after blind segmentation. The experimental results show that the training set accuracy is close to 100%, and the accuracy of the test set is more than 93%. It can recognize the license plate characters which cannot be recognized by the human eye.

Environmental sound recognition using short-time feature aggregation

Recognition of environmental sound is usually based on two main architectures, depending on whether the model is trained with frame-level features or with aggregated descriptions of acoustic scenes or events. The former architecture is appropriate for applications where target categories are known in advance, while the later affords a less supervised approach. In this paper, we propose a framework for environmental sound recognition based on blind segmentation and feature aggregation. We describe a new set of descriptors, based on Recurrence Quantification Analysis (RQA), which can be extracted from the similarity matrix of a time series of audio descriptors. We analyze their usefulness for recognition of acoustic scenes and events in addition to standard feature aggregation. Our results show the potential of non-linear time series analysis techniques for dealing with environmental sounds.

The Role of Semantic Context in Early Morphological Processing.

There is extensive evidence pointing to an early, automatic segmentation of written words into their constituent units (farm-er, wit-ness); however, less is known about the potential role of contextual information in modulating this analysis. We adapted the standard masked priming paradigm to include an overt semantic prime in order to examine whether semantic context influences morpho-orthographic segmentation of complex words. In particular, we asked how the context will affect processing of semantically opaque forms (witness), where the embedded stem (wit) is incompatible with the meaning of the whole form. Results showed no masked priming facilitation for opaque forms in the presence of a semantic prime, indicating that context can influence early morphological analysis. Priming was found for both semantically transparent and opaque forms (farmer-farm, witness-wit) when there was no semantically-related context, consistent with the literature and an account positing early blind segmentation. These findings provide an important update to the long-standing debate on early morphological processing in written word recognition.

基于显著性检测和改进投影字典对的盲道分割

基于显著性检测和改进投影字典对的盲道分割

HeartID: A Multiresolution Convolutional Neural Network for ECG-Based Biometric Human Identification in Smart Health Applications

Body area networks, including smart sensors, are widely reshaping health applications in the new era of smart cities. To meet increasing security and privacy requirements, physiological signalbased biometric human identification is gaining tremendous attention. This paper focuses on two major impediments: the signal processing technique is usually both complicated and data-dependent and the feature engineering is time-consuming and can fit only specific datasets . To enable a data-independent and highly generalizable signal processing and feature learning process, a novel wavelet domain multiresolution convolutional neural network is proposed. Specifically, it allows for blindly selecting a physiological signal segment for identification purpose, avoiding the complicated signal fiducial characteristics extraction process. To enrich the data representation, the random chosen signal segment is then transformed to the wavelet domain, where multiresolution time-frequency representation is achieved. An auto-correlation operation is applied to the transformed data to remove the phase difference as the result of the blind segmentation operation. Afterward, a multiresolution 1-D-convolutional neural network (1-D-CNN) is introduced to automatically learn the intrinsic hierarchical features from the wavelet domain raw data without datadependent and heavy feature engineering, and perform the user identification task. The effectiveness of the proposed algorithm is thoroughly evaluated on eight electrocardiogram datasets with diverse behaviors, such as with or without severe heart diseases, and with different sensor placement methods. Our evaluation is much more extensive than the state-of-the-art works, and an average identification rate of 93.5% is achieved. The proposed multiresolution 1-D-CNN algorithm can effectively identify human subjects, even from randomly selected signal segments and without heavy feature engineering. This paper is expected to demonstrate the feasibility and effectiveness of applying the blind signal processing and deep learning techniques to biometric human identification, to enable a low algorithm engineering effort and also a high generalization ability.

Non-parametric Bayesian inference for continuous density hidden Markov mixture model

In this paper, we present a non-parametric continuous density Hidden Markov mixture model (CDHMMix model) with unknown number of mixtures for blind segmentation or clustering of sequences. In our presented model, the emission distributions of HMMs are chosen to be Gaussian with full, diagonal, or tridiagonal covariance matrices. We apply a Bayesian approach to train our presented model and drive the inference of our model using the Monte Carlo Markov Chain (MCMC) method. For the multivariate Gaussian emission a method that maintains the tridiagonal structure of the covariance is introduced. Moreover, we present a new sampling method for hidden state sequences of HMMs based on the Viterbi algorithm that increases the mixing rate.

Integrated region-based segmentation using color components and texture features with prior shape knowledge

Integrated region-based segmentation using color components and texture features with prior shape knowledgeSegmentation is the art of partitioning an image into different regions where each one has some degree of uniformity in its feature space. A number of methods have been proposed and blind segmentation is one of them. It uses intrinsic image features, such as pixel intensity, color components and texture. However, some virtues, like poor contrast, noise and occlusion, can weaken the procedure. To overcome them, prior knowledge of the object of interest has to be incorporated in a top-down procedure for segmentation. Consequently, in this work, a novel integrated algorithm is proposed combining bottom-up (blind) and top-down (including shape prior) techniques. First, a color space transformation is performed. Then, an energy function (based on nonlinear diffusion of color components and directional derivatives) is defined. Next, signeddistance functions are generated from different shapes of the object of interest. Finally, a variational framework (based on the level set) is employed to minimize the energy function. The experimental results demonstrate a good performance of the proposed method compared with others and show its robustness in the presence of noise and occlusion. The proposed algorithm is applicable in outdoor and medical image segmentation and also in optical character recognition (OCR).

Speeding up HMRF_EM algorithms for fast unsupervised image segmentation by Bootstrap resampling: Application to the brain tissue segmentation

This work deals with global statistical unsupervised segmentation algorithms. In the context of Magnetic Resonance Image (MRI), an accurate and robust segmentation can be achieved by combining both the Hidden Markov Random Field (HMRF) model and the Expectation-Maximization (EM) algorithm. This EM–HMRF approach is accomplished by taking into account spatial information to improve the segmentation process which, in turn, slows the approach and consequently prevents its adoption for real-time applications such as three-dimensional medical image segmentation. We propose in this paper the use of the Bootstrap resampling to speed up the processing time of the EM–HMRF algorithm. This is accomplished by randomly selecting an optimal representative set of pixels according to some criteria originally defined for the blind segmentation. We will show how to adapt such criteria to the HMRF_EM algorithm context. We validated our proposition through a set of experiments and we proved that the use of the Bootstrap resampling yields the same accuracy and robustness as the basic algorithm, yet it amounts to a considerable processing speed up.

Morphology, language and the brain: the decompositional substrate for language comprehension

This paper outlines a neurocognitive approach to human language, focusing on inflectional morphology and grammatical function in English. Taking as a starting point the selective deficits for regular inflectional morphology of a group of non-fluent patients with left hemisphere damage, we argue for a core decompositional network linking left inferior frontal cortex with superior and middle temporal cortex, connected via the arcuate fasciculus. This network handles the processing of regularly inflected words (such as joined or treats), which are argued not to be stored as whole forms and which require morpho-phonological parsing in order to segment complex forms into stems and inflectional affixes. This parsing process operates early and automatically upon all potential inflected forms and is triggered by their surface phonological properties. The predictions of this model were confirmed in a further neuroimaging study, using event-related functional magnetic resonance imaging (fMRI), on unimpaired young adults. The salience of grammatical morphemes for the language system is highlighted by new research showing that similarly early and blind segmentation also operates for derivationally complex forms (such as darkness or rider). These findings are interpreted as evidence for a hidden decompositional substrate to human language processing and related to a functional architecture derived from non-human primate models.

Subword-based speaker verification with multiple-classifier score fusion weight and threshold adaptation

The voice print system of the present invention is a subword-based, text-dependent automatic speaker verification system that embodies the capability of user-selectable passwords with no constraints on the choice of vocabulary words or the language. Automatic blind speech segmentation allows speech to be segmented into subword units without any linguistic knowledge of the password. Subword modeling is performed using a multiple classifiers. The system also takes advantage of such concepts as multiple classifier fusion and data resampling to successfully boost the performance. Key word/key phrase spotting is used to optimally locate the password phrase. Numerous adaptation techniques increase the flexibility of the base system, and include: channel adaptation, fusion adaptation, model adaptation and threshold adaptation.

The patchwork engine: image segmentation from shape symmetries

We propose blind segmentation of images into shape-related ‘patches’ based on pre-calculated local symmetries (Van Tonder, G.J. & Ejima, Y. (1999). (Forthcoming a) Flexible computation of shape symmetries. Submitted for publication) in shape boundary contours. First, lateral weights between all points in the boundary contour map are assigned analogous to Euclidean distance maps in watershed segmentation (Beucher, S. & Lantejoul, C. (1979). Use of watersheds in contour detection. Proceedings of the International Workshop on Image Processing, CCETT, Rennes, France.). Lateral weights are then used to: (1) extract local maxima in symmetries; (2) link maxima within locally enclosed boundary contours; and (3) reconstruct shape contours using symmetry maxima as ‘seeds’. The new model overcomes weaknesses of watershed segmentation. The new model closes gaps in relatively more solid image contours, but it is fundamentally different from methods based on contour interpolation (Grossberg, S., Mingolla, E. & Todorovć, D. (1989). A neural network architecture for pre-attentive vision, IEEE Transactions on Biomedical Engineering 36, 65–84; Heitger, F. & von der Heydt, R. (1993). A computational model of neural contour processing: figure-ground segregation and illusory contours. Proceedings of the Fourth International Conference on Computer Vision, IEEE Computer Society Press, Washington D.C. (pp. 32–40)). Images are segmented into shape-relevant color-by-number-like patches which compare well to related methods (Gauch, J. & Pizer, M. (1993). The intensity axis of symmetry and its application to image segmentation, IEEE Transactions on Pattern Analysis and Machine Intelligence, 15 (8), 753–770; Ilg, W. & Ogniewicz, R. (1995). The application of Voronoi skeletons to perceptual grouping in line images, Proceedings of the 11th International Conference on Pattern Recognition, The Hague, The Netherlands, pp. 382–385; Zhu, S.C. & Yuille, A.L. (1996) FORMS: a flexible object recognition and modeling system, International Journal of Computer Vision, 20 (3), 187–212.). Two primitive operations, comparison and merging of patches, are proposed as drives for exposing more global shape contours from patches. We conclude that symmetry goes beyond abstract shape morphology: it can contribute to figure-ground segmentation in early vision and form part of primitive operations needed to create hypotheses of complex shape.

Estimation of fuzzy Gaussian mixture and unsupervised statistical image segmentation

This paper addresses the estimation of fuzzy Gaussian distribution mixture with applications to unsupervised statistical fuzzy image segmentation. In a general way, the fuzzy approach enriches the current statistical models by adding a fuzzy class, which has several interpretations in signal processing. One such interpretation in image segmentation is the simultaneous appearance of several thematic classes on the same site. We introduce a new procedure for estimating of fuzzy mixtures, which is an adaptation of the iterative conditional estimation (ICE) algorithm to the fuzzy framework, We first describe the blind estimation, i.e., without taking into account any spatial information, valid in any context of independent noisy observations. Then we introduce, in a manner analogous to classical hard segmentation, the spatial information by two different approaches: contextual segmentation and adaptive blind segmentation. In the first case, the spatial information is taken into account at the segmentation step level, and in the second case it is taken into account at the parameter estimation step level. The results obtained with the iterative conditional estimation algorithm are compared to those obtained with expectation-maximization (EM) and the stochastic EM algorithms, on both parameter estimation and unsupervised segmentation levels, via simulations. The methods proposed appear as complementary to the fuzzy C-means algorithms.

Fuzzy random fields and unsupervised image segmentation

Statistical unsupervised image segmentation using fuzzy random fields is treated. A fuzzy model containing a hard component, which describes pure pixels, and a fuzzy component which describes mixed pixels, is introduced. A procedure for simulating, a fuzzy field based on a Gibbs sampler step followed by a second step involving white or correlated Gaussian noises is given. Then the different steps of unsupervised image segmentation are studied. Four different blind segmentation methods are performed: the conditional expectation, two variants of the maximum likelihood, and the least squares approach. The parameters required are estimated by the stochastic estimation maximization (SEM) algorithm, a stochastic variant of the expectation maximization (EM) algorithm. These fuzzy segmentation methods are compared with a classical hard segmentation method, without taking the fuzzy class into account. The study shows that the fuzzy SEM algorithm provides reliables estimators. Furthermore, fuzzy segmentation always improves upon the hard segmentation results. >