Correct Recognition Rate Research Articles

Damage assessments of composite under the environment with strong noise based on synchrosqueezing wavelet transform and stack autoencoder algorithm

This study proposes a technique for damage location and quantitative identification for composites under strong noise background on the basis of synchro squeezing wavelet transform and stack autoencoder algorithm. Firstly, the simulation of the mechanism of Lamb wave and damage of different degrees were conduct in this stage. Secondly, the simulation of the actual damage was carried out in the selecting composites with strain field of the structure changed by different masses (within the range of 3 grades) and mass blocks at different locations (64 regions). Moreover, the white noise (3 dB) was introduced to the signal detected by the sensor in order to simulate the Lamb Wave signals collected from the sensors under the background of strong noise. Actually, the training sample is composed of 26,880 (3 × 64 × 140) signals. In addition, the synchronized wavelet transform was proposed to denoise the signal through employing strong noise. After that, Fourier Transform was applied to extract damaged frequencies. Additionally, the input of the stack autoencoder was the frequencies, and that output was the result of corresponding damage, and in the meanwhile the damage identification model was established as well. Finally, the test data which had been processed was then input into the damage identification model for further testing and analysis. It was finally concluded that 191(total 192) samples were correctly identified, and the correct recognition rate arrived as high as 99.48%. It can identify different degrees of damage in different locations. It was demonstrated by the results that the method is highly accurate, with extensive application potential in instantaneous locating and quantitative recogniting damages in the composite plate under the condition of strong noise exists.

Experimental Research on Construction-Waste-Classification Algorithms for a Hyperspectral-Camera System

A large proportion of construction waste has a high recovery value, and some of the existing recycling-classification methods rely mainly on physical properties for vibration screening. In order to effectively recover construction waste, an industrial near-infrared hyperspectral camera is proposed in this paper to distinguish the spectral characteristics of the objects. The testing results were verified using an extreme learning machine, an adaptive-learning multilayer perceptron, and a one-dimensional convolutional neural network classification model. By establishing several different models to classify and identify the same kinds of experimental materials, the experimental results not only output the correct recognition rate, but also use the recognition efficiency and stability as indices for comprehensive evaluation. The results show that different classification models have different efficiencies and levels of correctness. Under different analytical conditions, such as when using data from different bands, it is very important to select the appropriate classification model to classify construction waste.

Representation learning in a deep network for license plate recognition

The goal of license plate recognition (LPR) is to read the license plate characters. Due to image degradation, there are many difficulties in the way of achieving this goal. In this paper, the proposed method recognizes the license plate characters without employing the traditional segmentation and binarization techniques. This method uses a deep learning algorithm and tries to achieve better learning experience by engaging a multi-task learning algorithm based on sharing features. The features of license plate characters are extracted by a deep encoder-decoder network, and transferred to 8 parallel classifiers for recognition. To evaluate the current work, a database of 11,000 license plate images, collected from a currently working surveillance system installed on a dual carriageway, is employed. The proposed method achieved the correct character recognition rate of 96% for 4000 test images that is acceptable in comparison to the competing methods.

Selection of optimized features for fusion of palm print and finger knuckle‐based person authentication

AbstractThe impact of digital technology in biometrics is much more efficient at interpreting data than humans, which results in completely replacement of manual identification procedures in forensic science. Because the single modality‐based biometric frameworks limit performance in terms of accuracy and anti‐spoofing capabilities due to the presence of low quality data, therefore, information fusion of more than one biometric characteristic in pursuit of high recognition results can be beneficial. In this article, we present a multimodal biometric system based on information fusion of palm print and finger knuckle traits, which are least associated to any criminal investigation as evidence yet. The proposed multimodal biometric system might be useful to identify the suspects in case of physical beating or kidnapping and establish supportive scientific evidences, when no fingerprint or face information is present in photographs. The first step in our work is data preprocessing, in which region of interest of palm and finger knuckle images have been extracted. To minimize nonuniform illumination effects, we first normalize the detected circular palm or finger knuckle and then apply line ordinal pattern (LOP)‐based encoding scheme for texture enrichment. The nondecimated quaternion wavelet provides denser feature representation at multiple scales and orientations when extracted over proposed LOP encoding and increases the discrimination power of line and ridge features. To best of our knowledge, this first attempt is a combination of backtracking search algorithm and 2D2LDA has been employed to select the dominant palm and knuckle features for classification. The classifiers output for two modalities are combined at unsupervised rank level fusion rule through Borda count method, which shows an increase in performance in terms of recognition and verification, that is, 100% (correct recognition rate), 0.26% (equal error rate), 3.52 (discriminative index), and 1,262 m (speed).

Study on face recognition under unconstrained conditions based on LBP and deep learning

At present, the correct recognition rate of face recognition algorithm is limited under unconstrained conditions. To solve this problem, a face recognition algorithm based on deep learning under unconstrained conditions is proposed in this paper. The algorithm takes LBP texture feature as the input data of deep network, and trains the network layer by layer greedily to obtain optimized parameters of network, and then uses the trained network to predict the test samples. Experimental results on the face database LFW show that the proposed algorithm has higher correct recognition rate than some traditional algorithms under unconstrained conditions. In order to further verify its effectiveness and universality, this algorithm was also tested in YALE and YALE-B, and achieved a high correct recognition rate as well, which indicated that the deep learning method using LBP texture feature as input data is effective and robust to face recognition.

Hybrid model for Chinese character recognition based on Tesseract-OCR

Optical character recognition (OCR) is an important way to input information into a computer. And text information can be extracted by OCR from an image. Currently, the accuracy rate of Chinese OCR can also be improved. This study proposes a hybrid Chinese character recognition model based on the characteristics of Chinese. Before the OCR engine works, the model first filters the interference information in the image. Then the model adjusts the aspect ratio of the character. After an image is identified by OCR, single character recognition result is obtained. Then the result is checked and corrected on the phrase level. The experimental results show that the hybrid model improves the accuracy rate of Chinese OCR. Through image processing, the correct rate of recognition by the Tesseract-OCR engine is increased by about 12%, and the natural language processing improves the accuracy of the recognition result by about 5%.

Generic aircraft model recognition by two shape factors: in the resonance region

The study utilises two shape factors as the radar feature set for the recognition of a mid-sized aircraft target in the high-frequency (HF) band. The two shape factors are the dihedral and tilt features of the aircraft parts. The determination of the two shape factors is feasible through the optimum polarisation states of the target's complex natural resonances in the HF band. The identification algorithm applies a resonance-weighted measure to the feature set in a supervised learning approach to discriminate between two targets of the same class that slightly differ in shape. The rate of correct recognition within a sector of the target aspect exceeded 80% for receiver sensitivity above −10 dB.

A CNN-RBPNN Model With Feature Knowledge Embedding and its Application to Time-Varying Signal Classification

A novel technique, combining the feature extraction mechanisms of a convolutional neural network (CNN) with the classification method of a radial basis probability neural network (RBPNN), is proposed for small sample set modeling and feature knowledge embedding in multi-channel time-varying signal classification. This CNN-RBPNN consists of a signal input layer, signal feature parallel extraction and integration units, and an RBPNN classifier. Each channel signal in a feature extraction unit corresponds to a 1D CNN. The extracted features are represented as feature vectors, and these vectors constitute a comprehensive feature matrix. The RBPNN classifier was designed using signal feature embedding mechanism based on radial basis kernels and the property of combining pattern subclasses into pattern classes to form complex class boundaries. A dynamic clustering algorithm was used to divide each pattern class sample into several subclasses. Typical signal samples in each pattern subclass were designated as kernel centers, in order to achieve signal categories features embedding. This process was also used to determine the number of nodes in the RBPN layer. The RBPN layer outputs were selectively summed in the pattern layer according to kernel center category, which can generate irregular class boundaries, reducing the overlap among different pattern class boundary. The proposed CNN-RBPNN replaces the full-connection layer and classifier unit of conventional CNN with RBPNN, which can extract and represent signals distribution features and structural properties, implement structural and data constraints. This can reduce the structural risks of small sample set modeling. In this study, the properties of CNN-RBPNN are analyzed and an integrated learning algorithm is proposed. An experiment was conducted using 12-lead ECG signals in a seven-classification in the case of small sample set. Results demonstrated that, the correct recognition rate is 5.7% higher than other methods in the experiment, the performance evaluation index also showed significant improvement.

EEG-based Human Recognition Using Steady-State AEPs and Subject-Unique Spatial Filters

In recent years, brainwaves (EEG) have gained increasing attention in the field of biometric authentication because they feature vital advantages being more secure and impossible to replicate. In this paper, a new approach for the EEG-based biometric recognition system is proposed using steady-state Auditory Evoked Potentials (AEPs). This class of modular brainwaves adds extra features to the system like cancelability and two-step authentication. To investigate the biometric potential of AEPs, brainwaves from 40 subjects were recorded while being stimulated by multiple auditory tones modulated at two frequency bands; 40 Hz (m-40) and 80 Hz (m-80). Each subject participated in two sessions on two different days for time-permanence evaluation. Brain-Computer Interface (BCI) techniques were adopted here for the rapid estimation of the AEPs using canonical correlation analysis. The energy distribution of the AEPs in different frequency bands represented the subject-unique features. For intra-session setup, correct recognition rates up to 96.46% and equal error rates as low as 0% were achieved using the m-80 stimulation over all the 40 subjects. Moreover, results across different sessions showed high recognition rates (94.5 - 96.5%) and low error rates (2 - 4%) over the same number of subjects. These results show that AEPs carry subject discriminating features allowing the possibility of employing AEPs as a biometric trait.

A Novel Cancelable FaceHashing Technique Based on Non-Invertible Transformation With Encryption and Decryption Template

A novel cancelable FaceHashing technique based on non-invertible transformation with encryption and decryption template has been proposed in this paper. The proposed system has four components: face preprocessing, feature extraction, cancelable feature extraction followed by the classification, and encryption/decryption of cancelable face feature templates. During face preprocessing, the facial region of interest has been extracted out to speed the process for evaluating discriminant features. In feature extraction, some optimization techniques such as Sparse Representation Coding, Coordinate descent, and Block coordinates descent have been employed on facial descriptors to obtain the best representative of those descriptors. The representative descriptors are further arranged in a spatial pyramid matching structure to extract more discriminant and distinctive feature vectors. To preserve them, the existing BioHashing technique has been modified and extended to some higher levels of security attacks and the modified BioHashing technique computes a cancelable feature vector by the combined effect of the facial feature vector and the assigned token correspond to each user. The elements of computed cancelable feature vector are in a numeric form that has been employed to perform both verifications as well as identification task in online while the original facial feature vectors are kept offline either in hard drive or disc. Then, to enhance more security levels and also to preserve the cancelable face features, an RSA based encryption-decryption algorithm has been introduced. The proposed system has been tested using four benchmark face databases: CASIA-FACE-v5, IITK, CVL, and FERET, and performance are obtained as correct recognition rate and equal error rate. The performance are compared to the state-of-the-art methods for the superiority of the proposed feature extraction technique and individual performance analysis has been performed at all the security levels of the proposed Cancelable FaceHashing Technique. These comparisons show the superiority of the proposed system.

Two-phase palmprint identification

In this paper, a two-phase palmprint recognition approach is proposed based on statistical features and wide principal line image features through dynamic region of interest (ROI). The ROI is segmented into overlapping segments by six schemes, and the statistical features are extracted directly from the segments. The algorithm focuses on the extraction of statistical features based on standard deviation and coefficient of variation. A modified dissimilarity distance is proposed for computing the distance between two palmprints. The procedures are presented for determining the size and location of the common region of training images dynamically. Experiments are conducted by using statistical features and the combination of statistical and wide principal line image features. The results show that the correct recognition rate (CRR) of the proposed approach is better than existing methods for PolyUPalmprint database.

SVDTWDD Method for High Correct Recognition Rate Classifier With Appropriate Rejection Recognition Regions

At present, regions of the same class determined by Support Vector Machines (SVM) classifier, Support Vector Domain Description (SVDD) classifier and Deep Learning (DL) classifier may occupy regions of other classes or unknown classes in feature space. There exists a risk that samples of other classes or unknown classes are wrongly classified as a known class. In this paper, the Support Vector Domain Tightly Wrapping Description Design (SVDTWDD) method with appropriate rejection regions and the corresponding incremental learning algorithm are proposed to overcome the above problem. The main work includes: (1) We develop a construction algorithm of the tightly wrapping set for the homogeneous feature set; (2) Based on the homogeneous feature set and tightly wrapping set, a novel algorithm is presented for obtaining the tightly wrapping surface of the homogeneous feature region; (3) The method for constructing all the public regions outside of the tightly wrapping surface and the intersections of wrapping regions in two different tightly wrapping surfaces, as the rejection region of the classifier; (4) An incremental algorithm is also presented based on the SVD-TWDD method. The experimental results with UCI data sets show that the correct recognition rate of our proposed method is nearly100% even if with a low rejection rate.

Palmprint Recognition using Robust Template Matching

Palmprint recognition system is extensively deployed in a variety of applications, ranging from forensic to mobile phones. This paper proposed a non-invertible palmprint template that stores the relative geometric information of the minutiae points. The information about the minutiae orientation and their location are not available in the template, which makes impossible to estimate the orientation of the palmprint from the template. The proposed template is robust to rotation and scaling and immune to reconstruction algorithm. Finally, Delaunay triangulation based internal angle matching which is computationally efficient is used for template matching. The proposed method is tested on standard PolyU, IIT-Delhi and Multi-spectral palm-print databases and yields better results in terms of Correct Recognition Rate (CRR) and Equal Error Rate (EER) of 95.4% and 0.37% respectively.

The Analysis of Plants Image Recognition Based on Deep Learning and Artificial Neural Network

Classification and identification of plants are helpful for people to effectively understand and protect plants. The leaves of plants are the most important recognition organs. With the development of artificial intelligence and machine vision technology, plant leaf recognition technology based on image analysis is used to improve the knowledge of plant classification and protection. Deep learning is the abbreviation of deep neural network learning method and belongs to neural network structure. It can automatically learn features from big data and use artificial neural network based on back propagation algorithm to train and classify plant leaf samples. The main content of this paper is to extract plant leaf features and identify plant species based on image analysis. Firstly, plant leaf images are segmented by various methods, and then feature extraction algorithm is used to extract leaf shape and texture features from leaf sample images. Then the comprehensive characteristic information of plant leaves is formed according to the comprehensive characteristic information. In this paper, 50 plant leaf databases are tested and compared with KNN-based neighborhood classification, Kohonen network based on self-organizing feature mapping algorithm and SVM-based support vector machine. At the same time, the leaves of 7 different plants were compared and it was found that ginkgo leaves were easier to identify. For leaf images under complex background, good recognition effect has been achieved. Image samples of the test set are input into the learning model to obtain reconstruction errors. The class label of the test set can be obtained by reconstructing the deep learning model with the smallest error set. The results show that this method has the shortest recognition time and the highest correct recognition rate.

Wireless Channel Scene Recognition Method Based on an Autocorrelation Function and Deep Learning

Wireless channel scene recognition plays a key role in cognitive radio (CR) mobile communication systems. This paper proposes a wireless channel scene identification framework based on the autocorrelation function and deep learning. First, a feature extraction (FE) method is developed to perform a channel scene date analysis based on the autocorrelation function (AF). The AF is used to realize the FE method because it can be combined with Fourier transform (FT) to accurately extract the characteristics accurately from a time-varying channels scene. Second, a deep belief network (DBN) with a robust learning approach is introduced to perform wireless channel scene recognition. A novel learning architecture is employed, which combines the feature parameter and classification techniques to achieve a high classification correct recognition rate. Third, the k-step contrastive divergence (CD-k) algorithm is introduced as the preliminary training method to optimize the traditional DBN network. This method can effectively calculate the logarithmic gradient of the Boltzmann machine. Moreover, the up-down optimization algorithm is applied to optimize the network parameters. Finally, the theoretical implementation is described in detail, and the method is verified by constructing an experiment platform for an engineering application. The experimental results indicate that the proposed classifier is an excellent approach to realize channel scene recognition through advanced methods. The classification accuracy of the proposed approach is higher than that of several existing techniques.

Automatic Identification of Breast Ultrasound Image Based on Supervised Block-Based Region Segmentation Algorithm and Features Combination Migration Deep Learning Model.

Breast cancer is a high-incidence type of cancer for women. Early diagnosis plays a crucial role in the successful treatment of the disease and the effective reduction of deaths. In this paper, deep learning technology combined with ultrasound imaging diagnosis was used to identify and determine whether the tumors were benign or malignant. First, the tumor regions were segmented from the breast ultrasound (BUS) images using the supervised block-based region segmentation algorithm. Then, a VGG-19 network pretrained on the ImageNet dataset was applied to the segmented BUS images to predict whether the breast tumor was benign or malignant. The benchmark data for bio-validation were obtained from 141 patients with 199 breast tumors, including 69 cases of malignancy and 130 cases of benign tumors. The experiment showed that the accuracy of the supervised block-based region segmentation algorithm was almost the same as that of manual segmentation; therefore, it can replace manual work. The diagnostic effect of the combination feature model established based on the depth feature of the B-mode ultrasonic imaging and strain elastography was better than that of the model established based on these two images alone. The correct recognition rate was 92.95%, and the AUC was 0.98 for the combination feature model.

Evaluation method for HRRP imaging quality based on intuitionistic interval fuzzy set

Aimed at the performance evaluation of radar target recognition system based on high resolution range profile (HRRP), this paper proposes an evaluation method of HRRP quality based on intuitionistic interval fuzzy set under the background that HRRP has the characteristics of translation, attitude and velocity sensitivity. Compared with the general evaluation method, this method has two improvements. One is to extend the index value from single value to interval value by interval estimation method in mathematical statistics, which introduces sample variance into calculation to improve the utilization rate of sample information. The other is to construct intuitionistic fuzzy set by using correct recognition rate derivative sequence (DSCRR) and false recognition rate derivative sequence (DSFRR), which can make full use of experimental data and improve the reliability of experimental results. The distance calculation method of intuitionistic interval fuzzy set is used to get the distance between each scheme and the ideal scheme. Finally, the score of schemes which represent different HRRP imaging quality are obtained, and the performance evaluation is completed. The evaluation results obtained from simulation validation experiments are in accordance with objective facts, which proves the validity of this evaluation method.

Gait feature extraction and gait classification using two-branch CNN

As a promising biometric identification method, gait recognition has many advantages, such as suitable for human identification at a long distance, requiring no contact and hard to imitate. However, due to the complex external factors in the gait data sampling process and the clothing changes of the person to be identified, gait recognition still faces numerous challenges in practical applications. In this paper, we present a novel solution for gait feature extraction and gait classification. Firstly, two kinds of Two-branch Convolution Neural Network (TCNN), i.e., middle-fusion TCNN and last-fusion TCNN, to improve the correct recognition rate of gait recognition are presented. Secondly, we construct Multi-Frequency Gait Energy Images (MF-GEIs) to train the proposed TCNNs networks and then extract refined gait features using the trained TCNNs. Finally, the output of each TCNN is utilized to train an SVM gait classifier separately which will be used for gait classification and recognition. In addition, the proposed solution is measured on CASIA dataset B and OU-ISIR LP dataset. Both experimental results show that our solution outperforms various existing methods.

Individual Identification Based on Code-Modulated Visual-Evoked Potentials

The electroencephalography (EEG) method has recently attracted increasing attention in the study of brain activity-based biometric systems because of its simplicity, portability, noninvasiveness, and relatively low cost. However, due to the low signal-to-noise ratio of EEG, most of the existing EEG-based biometric systems require a long duration of signals to achieve high accuracy in individual identification. Besides, the feasibility and stability of these systems have not yet been conclusively reported, since most studies did not perform longitudinal evaluation. In this paper, we proposed a novel EEG-based individual identification method using code-modulated visual-evoked potentials (c-VEPs). Specifically, this paper quantitatively compared eight code-modulated stimulation patterns, including six 63-bit (1.05 s at 60-Hz refresh rate) m-sequences (M1–M6) and two spatially combined sequence groups ( $\text{M}\times \text{4}$ : M1–M4 and $\text{M}\times \text{6}$ : M1–M6) in recording the c-VEPs from a group of 25 subjects for individual identification. To further evaluate the influence of inter-session variability, we recorded two data sessions for each individual on different days to measure intra-session and cross-session identification performance. State-of-the-art VEP detection algorithms in brain–computer interfaces (BCIs) were employed to construct a template-matching-based identification framework. For intra-session identification, we achieved a 100% correct recognition rate (CRR) using 5.25-s EEG data (average of five trials for M5). For cross-session identification, 99.43% CRR was attained using 10.5-s EEG signals (average of ten trials for M5). These results suggest that the proposed c-VEP-based individual identification method is promising for real-world applications.

Machine Learning-Based Classification of the Health State of Mice Colon in Cancer Study from Confocal Laser Endomicroscopy

In this article, we address the problem of the classification of the health state of the colon’s wall of mice, possibly injured by cancer with machine learning approaches. This problem is essential for translational research on cancer and is a priori challenging since the amount of data is usually limited in all preclinical studies for practical and ethical reasons. Three states considered including cancer, health, and inflammatory on tissues. Fully automated machine learning-based methods are proposed, including deep learning, transfer learning, and shallow learning with SVM. These methods addressed different training strategies corresponding to clinical questions such as the automatic clinical state prediction on unseen data using a pre-trained model, or in an alternative setting, real-time estimation of the clinical state of individual tissue samples during the examination. Experimental results show the best performance of 99.93% correct recognition rate obtained for the second strategy as well as the performance of 98.49% which were achieved for the more difficult first case.