Original Training Data Research Articles

Voice Conversion Based Augmentation and a Hybrid CNN-LSTM Model for Improving Speaker-Independent Keyword Recognition on Limited Datasets

Keyword recognition is the basis of speech recognition, and its application is rapidly increasing in keyword spotting, robotics, and smart home surveillance. Because of these advanced applications, improving the accuracy of keyword recognition is crucial. In this paper, we proposed voice conversion (VC) - based augmentation to increase the limited training dataset and a fusion of a convolutional neural network (CNN) and long-short term memory (LSTM) model for robust speaker-independent isolated keyword recognition. Collecting and preparing a sufficient amount of voice data for speaker-independent speech recognition is a tedious and bulky task. To overcome this, we generated new raw voices from the original voices using an auxiliary classifier conditional variational autoencoder (ACVAE) method. In this study, the main intention of voice conversion is to obtain numerous and various human-like keywords’ voices that are not identical to the source and target speakers’ pronunciation. Parallel VC was used to accurately maintain the linguistic content. We examined the performance of the proposed voice conversion augmentation techniques using robust deep neural network algorithms. Original training data, excluding generated voice using other data augmentation and regularization techniques, were considered as the baseline. The results showed that incorporating voice conversion augmentation into the baseline augmentation techniques and applying the CNN-LSTM model improved the accuracy of isolated keyword recognition.

A Mutual Guide Framework for Training Hyperspectral Image Classifiers With Small Data

This article develops a general yet effective hyperspectral image (HSI) classification framework that is trained with small data. To this end, two identically structured but differently initialized classifiers, which are referred to as two base classifiers, are trained in an iterative manner. Each iteration consists of three steps, that is: 1) the two base classifiers that are trained separately on guide data; 2) unclassified data that are processed by the two trained base classifiers; and 3) the classification results with high confidence that are explored as new guide data. In the first iteration, the guide data comprising the original small training data are the same for the two base classifiers. From the second iteration, the guide data for the two base classifiers start becoming different. Specifically, in each iteration, the guide data for training one base classifier keep being augmented by high confidence classification results provided by the other base classifier. It is in such an iterative manner that the two classifiers continuously provide different new guide data for each other, and thus increasingly augment labeled data from the original small training set to a reasonably larger amount of samples in a HSI. We refer to such a training strategy as mutual guide. We develop a mutual guide implementation scheme by exploiting extreme learning machines (ELMs) as base classifiers. Extensive experiments on four public HSI datasets, i.e., Indian Pines (IP), Kennedy Space Center (KSC), University of Pavia (UP), and Salinas (SA), validate the classification effectiveness of our mutual guide framework with small training data.

Generative Adversarial Minority Oversampling for Spectral–Spatial Hyperspectral Image Classification

Recently, convolutional neural networks (CNNs) have exhibited commendable performance for hyperspectral image (HSI) classification. Generally, an important number of samples are needed for each class to properly train CNNs. However, existing HSI data sets suffer from a significant class imbalance problem, where many classes do not have enough samples to characterize the spectral information. The performance of existing CNN models is biased toward the majority classes, which possess more samples for the training. This article addresses this issue of imbalanced data in HSI classification. In particular, a new <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3D-HyperGAMO</monospace> model is proposed, which uses generative adversarial minority oversampling. The proposed <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3D-HyperGAMO</monospace> automatically generates more samples for minority classes at training time, using the existing samples of that class. The samples are generated in the form of a 3-D hyperspectral patch. A different classifier from the generator and the discriminator is used in the <monospace xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3D-HyperGAMO</monospace> model, which is trained using both original and generated samples to determine the classes of newly generated samples to which they actually belong. The generated data are combined classwise with the original training data set to learn the network parameters of the class. Finally, the trained 3-D classifier network validates the performance of the model using the test set. Four benchmark HSI data sets, namely, Indian Pines (IP), Kennedy Space Center (KSC), University of Pavia (UP), and Botswana (BW), have been considered in our experiments. The proposed model shows outstanding data generation ability during the training, which significantly improves the classification performance over the considered data sets. The source code is available publicly at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/mhaut/3D-HyperGAMO</uri> .

Conditional Generative Data-Free Knowledge Distillation

Knowledge distillation has made remarkable achievements in model compression. However, most existing methods demand original training data, while real data in practice are often unavailable due to privacy, security and transmission limitation. To address this problem, we propose a conditional generative data-free knowledge distillation (CGDD) framework to train efficient portable network without any real data. In this framework, except using the knowledge extracted from teacher model, we introduce preset labels as additional auxiliary information to train the generator. Then, the trained generator can produce meaningful training samples of specified category as required. In order to promote distillation process, except using conventional distillation loss, we treat preset label as ground truth label so that student network is directly supervised by the category of synthetic training sample. Moreover, we force student network to mimic the attention maps of teacher model and further improve its performance. To verify the superiority of our method, we design a new evaluation metric is called as relative accuracy to directly compare the effectiveness of different distillation methods. Trained portable network learned with proposed data-free distillation method obtains 99.63%, 99.07% and 99.84% relative accuracy on CIFAR10, CIFAR100 and Caltech101, respectively. The experimental results demonstrate the superiority of proposed method.

MODIFIED CORRELATION WEIGHT K-NEAREST NEIGHBOR CLASSIFIER USING TRAINING DATASET CLEANING METHOD

In pattern recognition, the k-Nearest Neighbor (kNN) algorithm is the simplest non-parametric algorithm. Due to its simplicity, the model cases and the quality of the training data itself usually influence kNN algorithm classification performance. Therefore, this article proposes a sparse correlation weight model, combined with the Training Data Set Cleaning (TDC) method by Classification Ability Ranking (CAR) called the CAR classification method based on Coefficient-Weighted kNN (CAR-CWKNN) to improve kNN classifier performance. Correlation weight in Sparse Representation (SR) has been proven can increase classification accuracy. The SR can show the 'neighborhood' structure of the data, which is why it is very suitable for classification based on the Nearest Neighbor. The Classification Ability (CA) function is applied to classify the best training sample data based on rank in the cleaning stage. The Leave One Out (LV1) concept in the CA works by cleaning data that is considered likely to have the wrong classification results from the original training data, thereby reducing the influence of the training sample data quality on the kNN classification performance. The results of experiments with four public UCI data sets related to classification problems show that the CAR-CWKNN method provides better performance in terms of accuracy.

Combining Self-supervised Learning and Active Learning for Disfluency Detection

Spoken language is fundamentally different from the written language in that it contains frequent disfluencies or parts of an utterance that are corrected by the speaker. Disfluency detection (removing these disfluencies) is desirable to clean the input for use in downstream NLP tasks. Most existing approaches to disfluency detection heavily rely on human-annotated data, which is scarce and expensive to obtain in practice. To tackle the training data bottleneck, in this work, we investigate methods for combining self-supervised learning and active learning for disfluency detection. First, we construct large-scale pseudo training data by randomly adding or deleting words from unlabeled data and propose two self-supervised pre-training tasks: (i) a tagging task to detect the added noisy words and (ii) sentence classification to distinguish original sentences from grammatically incorrect sentences. We then combine these two tasks to jointly pre-train a neural network. The pre-trained neural network is then fine-tuned using human-annotated disfluency detection training data. The self-supervised learning method can capture task-special knowledge for disfluency detection and achieve better performance when fine-tuning on a small annotated dataset compared to other supervised methods. However, limited in that the pseudo training data are generated based on simple heuristics and cannot fully cover all the disfluency patterns, there is still a performance gap compared to the supervised models trained on the full training dataset. We further explore how to bridge the performance gap by integrating active learning during the fine-tuning process. Active learning strives to reduce annotation costs by choosing the most critical examples to label and can address the weakness of self-supervised learning with a small annotated dataset. We show that by combining self-supervised learning with active learning, our model is able to match state-of-the-art performance with just about 10% of the original training data on both the commonly used English Switchboard test set and a set of in-house annotated Chinese data.

Polymorphic Adversarial Cyberattacks Using WGAN

Intrusion Detection Systems (IDS) are essential components in preventing malicious traffic from penetrating networks and systems. Recently, these systems have been enhancing their detection ability using machine learning algorithms. This development also forces attackers to look for new methods for evading these advanced Intrusion Detection Systemss. Polymorphic attacks are among potential candidates that can bypass the pattern matching detection systems. To alleviate the danger of polymorphic attacks, the IDS must be trained with datasets that include these attacks. Generative Adversarial Network (GAN) is a method proven in generating adversarial data in the domain of multimedia processing, text, and voice, and can produce a high volume of test data that is indistinguishable from the original training data. In this paper, we propose a model to generate adversarial attacks using Wasserstein GAN (WGAN). The attack data synthesized using the proposed model can be used to train an IDS. To evaluate the trained IDS, we study several techniques for updating the attack feature profile for the generation of polymorphic data. Our results show that by continuously changing the attack profiles, defensive systems that use incremental learning will still be vulnerable to new attacks; meanwhile, their detection rates improve incrementally until the polymorphic attack exhausts its profile variables.

Applying Data Augmentation and Mask R-CNN-Based Instance Segmentation Method for Mixed-Type Wafer Maps Defect Patterns Classification

Defect patterns on semiconductor wafer maps point to different manufacturing problems. Consequently, they have become key factors in identifying and resolving the root causes of yield improvement. Perhaps not surprisingly, the probability of having multiple defect patterns on a wafer map has increased in tandem with advances in manufacturing technology. Prior research on defect patterns has focused primarily on image classification methods which are neither good at mixed-type defect pattern classification nor able to provide locational information for further analysis. This study develops a framework that integrates a Mask R-CNN-based instance segmentation model with copy-paste and rotational data augmentation. The proposed method is able to precisely classify and locate defect patterns on a wafer map given limited training data, tasks which can help companies identify the manufacturing root causes of defects in a timely manner when ramping up their production for yield enhancement. Our experiments were performed using real-world WM-811K data. Using COCO pre-trained weights and only 1,056 items of original training data, the model reached an accuracy of 97.7% on single-type classification. Mixed-type classification hamming loss, exact match and accuracy were 0.155, 69% and 82%, respectively.

Black-Box Diagnosis and Calibration on GAN Intra-Mode Collapse: A Pilot Study

Generative adversarial networks (GANs) nowadays are capable of producing images of incredible realism. Two concerns raised are whether the state-of-the-art GAN’s learned distribution still suffers from mode collapse and what to do if so. Existing diversity tests of samples from GANs are usually conducted qualitatively on a small scale and/or depend on the access to original training data as well as the trained model parameters. This article explores GAN intra-mode collapse and calibrates that in a novel black-box setting: access to neither training data nor the trained model parameters is assumed. The new setting is practically demanded yet rarely explored and significantly more challenging. As a first stab, we devise a set of statistical tools based on sampling that can visualize, quantify, and rectify intra-mode collapse . We demonstrate the effectiveness of our proposed diagnosis and calibration techniques, via extensive simulations and experiments, on unconditional GAN image generation (e.g., face and vehicle). Our study reveals that the intra-mode collapse is still a prevailing problem in state-of-the-art GANs and the mode collapse is diagnosable and calibratable in black-box settings. Our codes are available at https://github.com/VITA-Group/BlackBoxGANCollapse .

Dual discriminator adversarial distillation for data-free model compression

Knowledge distillation has been widely used to produce portable and efficient neural networks which can be well applied on edge devices for computer vision tasks. However, almost all top-performing knowledge distillation methods need to access the original training data, which usually has a huge size and is often unavailable. To tackle this problem, we propose a novel data-free approach in this paper, named Dual Discriminator Adversarial Distillation (DDAD) to distill a neural network without the need of any training data or meta-data. To be specific, we use a generator to create samples through dual discriminator adversarial distillation, which mimics the original training data. The generator not only uses the pre-trained teacher’s intrinsic statistics in existing batch normalization layers but also obtains the maximum discrepancy from the student model. Then the generated samples are used to train the compact student network under the supervision of the teacher. The proposed method obtains an efficient student network which closely approximates its teacher network, without using the original training data. Extensive experiments are conducted to demonstrate the effectiveness of the proposed approach on CIFAR, Caltech101 and ImageNet datasets for classification tasks. Moreover, we extend our method to semantic segmentation tasks on several public datasets such as CamVid, NYUv2, Cityscapes and VOC 2012. To the best of our knowledge, this is the first work on generative model based data-free knowledge distillation on large-scale datasets such as ImageNet, Cityscapes and VOC 2012. Experiments show that our method outperforms all baselines for data-free knowledge distillation.

Convolutional-network models to predict wall-bounded turbulence from wall quantities

Two models based on convolutional neural networks are trained to predict the two-dimensional instantaneous velocity-fluctuation fields at different wall-normal locations in a turbulent open-channel flow, using the wall-shear-stress components and the wall pressure as inputs. The first model is a fully convolutional neural network (FCN) which directly predicts the fluctuations, while the second one reconstructs the flow fields using a linear combination of orthonormal basis functions, obtained through proper orthogonal decomposition (POD), and is hence named FCN-POD. Both models are trained using data from direct numerical simulations at friction Reynolds numbers $Re_{\tau } = 180$ and 550. Being able to predict the nonlinear interactions in the flow, both models show better predictions than the extended proper orthogonal decomposition (EPOD), which establishes a linear relation between the input and output fields. The performance of the models is compared based on predictions of the instantaneous fluctuation fields, turbulence statistics and power-spectral densities. FCN exhibits the best predictions closer to the wall, whereas FCN-POD provides better predictions at larger wall-normal distances. We also assessed the feasibility of transfer learning for the FCN model, using the model parameters learned from the $Re_{\tau }=180$ dataset to initialize those of the model that is trained on the $Re_{\tau }=550$ dataset. After training the initialized model at the new $Re_{\tau }$ , our results indicate the possibility of matching the reference-model performance up to $y^{+}=50$ , with $50\,\%$ and $25\,\%$ of the original training data. We expect that these non-intrusive sensing models will play an important role in applications related to closed-loop control of wall-bounded turbulence.

An update to the custom-made MS Excel workbook performing the log-rank test with extended functionality and a new original COVID-19 training data set

An update to the custom-made MS Excel workbook performing the log-rank test with extended functionality and a new original COVID-19 training data set

Deep Learning with Data Augmentation to Add Data Around Classification Boundaries

Data augmentation methods are used as a technique to improve generalization by increasing the number of training data in image classification. However, most of these methods are not a data driven algorithm, the degree of improvement of generalization ability by performing these data augmentation methods differs between the domains of image data for training. Generative models are researched to use for augmenting data recently. In particular, Generative Adversarial Networks (GANs) (Goodfellow et al., 2014) that can generate clean image get attention as an excellent innovation in machine learning. As GANs extension method, there is a method called CGANs (Mirza and Osindero, 2014) that can be used for data augmentation. When enough training data for each class are not prepared for classification model, the same is true for training CGANs. In such case, CGAN generates noisy images. This makes a classification model to underfit to the original training data. Moreover, when a CGAN approximates the training data distribution, the CGAN generates new training data in the same region where training data densely exist. In such case, augmented data can’t reduce overfitting on the original training data. Therefore, our research contributes to augment data which meets these two requirements. In this study, we propose a method to generate data by the class specific GAN with small training data and selectively add generated data to the training data set that improves classification accuracy by using the entropy of the classification model. The feature of the proposed method is that it focuses on the positional relationship between data and the classification hyperplane in deep learning. In the proposed method, the entropy of the classification model is used to measure the positional relationship between the classification boundary and the data. As a result, the generalization performance is improved by adding the data around the classification boundary as new training data.

An ensemble machine learning model based on multiple filtering and supervised attribute clustering algorithm for classifying cancer samples.

BackgroundMachine learning is one kind of machine intelligence technique that learns from data and detects inherent patterns from large, complex datasets. Due to this capability, machine learning techniques are widely used in medical applications, especially where large-scale genomic and proteomic data are used. Cancer classification based on bio-molecular profiling data is a very important topic for medical applications since it improves the diagnostic accuracy of cancer and enables a successful culmination of cancer treatments. Hence, machine learning techniques are widely used in cancer detection and prognosis.MethodsIn this article, a new ensemble machine learning classification model named Multiple Filtering and Supervised Attribute Clustering algorithm based Ensemble Classification model (MFSAC-EC) is proposed which can handle class imbalance problem and high dimensionality of microarray datasets. This model first generates a number of bootstrapped datasets from the original training data where the oversampling procedure is applied to handle the class imbalance problem. The proposed MFSAC method is then applied to each of these bootstrapped datasets to generate sub-datasets, each of which contains a subset of the most relevant/informative attributes of the original dataset. The MFSAC method is a feature selection technique combining multiple filters with a new supervised attribute clustering algorithm. Then for every sub-dataset, a base classifier is constructed separately, and finally, the predictive accuracy of these base classifiers is combined using the majority voting technique forming the MFSAC-based ensemble classifier. Also, a number of most informative attributes are selected as important features based on their frequency of occurrence in these sub-datasets.ResultsTo assess the performance of the proposed MFSAC-EC model, it is applied on different high-dimensional microarray gene expression datasets for cancer sample classification. The proposed model is compared with well-known existing models to establish its effectiveness with respect to other models. From the experimental results, it has been found that the generalization performance/testing accuracy of the proposed classifier is significantly better compared to other well-known existing models. Apart from that, it has been also found that the proposed model can identify many important attributes/biomarker genes.

Privacy preservation in federated learning: An insightful survey from the GDPR perspective

In recent years, along with the blooming of Machine Learning (ML)-based applications and services, ensuring data privacy and security have become a critical obligation. ML-based service providers not only confront with difficulties in collecting and managing data across heterogeneous sources but also challenges of complying with rigorous data protection regulations such as EU/UK General Data Protection Regulation (GDPR). Furthermore, conventional centralised ML approaches have always come with long-standing privacy risks to personal data leakage, misuse, and abuse. Federated learning (FL) has emerged as a prospective solution that facilitates distributed collaborative learning without disclosing original training data. Unfortunately, retaining data and computation on-device as in FL are not sufficient for privacy-guarantee because model parameters exchanged among participants conceal sensitive information that can be exploited in privacy attacks. Consequently, FL-based systems are not naturally compliant with the GDPR. This article is dedicated to surveying of state-of-the-art privacy-preservation techniques in FL in relations with GDPR requirements. Furthermore, insights into the existing challenges are examined along with the prospective approaches following the GDPR regulatory guidelines that FL-based systems shall implement to fully comply with the GDPR.

Applications of object detection in modular construction based on a comparative evaluation of deep learning algorithms

Purpose The practice of artificial intelligence (AI) is increasingly being promoted by technology developers. However, its adoption rate is still reported as low in the construction industry due to a lack of expertise and the limited reliable applications for AI technology. Hence, this paper aims to present the detailed outcome of experimentations evaluating the applicability and the performance of AI object detection algorithms for construction modular object detection. Design/methodology/approach This paper provides a thorough evaluation of two deep learning algorithms for object detection, including the faster region-based convolutional neural network (faster RCNN) and single shot multi-box detector (SSD). Two types of metrics are also presented; first, the average recall and mean average precision by image pixels; second, the recall and precision by counting. To conduct the experiments using the selected algorithms, four infrastructure and building construction sites are chosen to collect the required data, including a total of 990 images of three different but common modular objects, including modular panels, safety barricades and site fences. Findings The results of the comprehensive evaluation of the algorithms show that the performance of faster RCNN and SSD depends on the context that detection occurs. Indeed, surrounding objects and the backgrounds of the objects affect the level of accuracy obtained from the AI analysis and may particularly effect precision and recall. The analysis of loss lines shows that the loss lines for selected objects depend on both their geometry and the image background. The results on selected objects show that faster RCNN offers higher accuracy than SSD for detection of selected objects. Research limitations/implications The results show that modular object detection is crucial in construction for the achievement of the required information for project quality and safety objectives. The detection process can significantly improve monitoring object installation progress in an accurate and machine-based manner avoiding human errors. The results of this paper are limited to three construction sites, but future investigations can cover more tasks or objects from different construction sites in a fully automated manner. Originality/value This paper’s originality lies in offering new AI applications in modular construction, using a large first-hand data set collected from three construction sites. Furthermore, the paper presents the scientific evaluation results of implementing recent object detection algorithms across a set of extended metrics using the original training and validation data sets to improve the generalisability of the experimentation. This paper also provides the practitioners and scholars with a workflow on AI applications in the modular context and the first-hand referencing data.

Multilingual Transfer Learning for QA using Translation as Data Augmentation

Prior work on multilingual question answering has mostly focused on using large multilingual pre-trained language models (LM) to perform zero-shot language-wise learning: train a QA model on English and test on other languages. In this work, we explore strategies that improve cross-lingual transfer by bringing the multilingual embeddings closer in the semantic space. Our first strategy augments the original English training data with machine translation-generated data. This results in a corpus of multilingual silver-labeled QA pairs that is 14 times larger than the original training set. In addition, we propose two novel strategies, language adversarial training and language arbitration framework, which significantly improve the (zero-resource) cross-lingual transfer performance and result in LM embeddings that are less language-variant. Empirically, we show that the proposed models outperform the previous zero-shot baseline on the recently introduced multilingual MLQA and TyDiQA datasets.

Robustness to Spurious Correlations in Text Classification via Automatically Generated Counterfactuals

Spurious correlations threaten the validity of statistical classifiers. While model accuracy may appear high when the test data is from the same distribution as the training data, it can quickly degrade when the test distribution changes. For example, it has been shown that classifiers perform poorly when humans make minor modifications to change the label of an example. One solution to increase model reliability and generalizability is to identify causal associations between features and classes. In this paper, we propose to train a robust text classifier by augmenting the training data with automatically generated counterfactual data. We first identify likely causal features using a statistical matching approach. Next, we generate counterfactual samples for the original training data by substituting causal features with their antonyms and then assigning opposite labels to the counterfactual samples. Finally, we combine the original data and counterfactual data to train a robust classifier. Experiments on two classification tasks show that a traditional classifier trained on the original data does very poorly on human-generated counterfactual samples (e.g., 10%-37% drop in accuracy). However, the classifier trained on the combined data is more robust and performs well on both the original test data and the counterfactual test data (e.g., 12%-25% increase in accuracy compared with the traditional classifier). Detailed analysis shows that the robust classifier makes meaningful and trustworthy predictions by emphasizing causal features and de-emphasizing non-causal features.

Continual Learning for Named Entity Recognition

Named Entity Recognition (NER) is a vital task in various NLP applications. However, in many real-world scenarios (e.g., voice-enabled assistants) new named entities are frequently introduced, entailing re-training NER models to support these new entities. Re-annotating the original training data for the new entities could be costly or even impossible when storage limitations or security concerns restrict access to that data, and annotating a new dataset for all of the entities becomes impractical and error-prone as the number of entities increases. To tackle this problem, we introduce a novel Continual Learning approach for NER, which requires new training material to be annotated only for the new entities. To preserve the existing knowledge previously learned by the model, we exploit the Knowledge Distillation (KD) framework, where the existing NER model acts as the teacher for a new NER model (i.e., the student), which learns the new entity by using the new training material and retains knowledge of old entities by imitating the teacher's outputs on this new training set. Our experiments show that this approach allows the student model to ``progressively'' learn to identify new entities without forgetting the previously learned ones. We also present a comparison with multiple strong baselines to demonstrate that our approach is superior for continually updating an NER model.

An Improved Discriminative Model Prediction Approach to Real-Time Tracking of Objects With Camera as Sensors

Generic person tracking is a basic task in visual surveillance by using camera as sensors. Many deep learning-based trackers have obtained outstanding performance. Among them, trackers based on Siamese networks have drawn great attention and are promising. These training methods are competitive but training data can be more effectively augmented to improve their person tracking performance. Many other trackers use only one layer to extract semantic features, likely hindering their discriminative learning. In this paper, we propose an enhanced discriminative model prediction method with efficient data augmentation and robust feature fusion. Specifically, we propose to implement an effective data augmentation strategy (e.g., color jitter and motion blur) to unleash the greater potential of original training data. We also adopt a multi-layer feature fusion to obtain a more discriminative feature map. Thus, the proposed tracker can discriminate an object in complicated scenarios in real time. We conduct extensive experiments on two datasets, i.e., VOT2018 and UAV123. Objective evaluation on VOT2018 demonstrates that with its expected average overlap value of 0.430, it outperforms a state-of-the-art tracker by 4.88%. On UAV123, it does so by 4.5% in success rate and 4.4% in precision rate. In addition, our further experimental results reveal that our algorithm can reach a speed that is high enough to meet the real-time tracking requirement when camera are used as sensors.