International Journal of Biometrics (33 papers in press)
Sub-band based Feature Fusion and Hybrid Fusion Approaches for Multimodal Biometric Identification
by Rajeshwari Devi D V
Abstract: A Multimodal biometric system using feature fusion and hybrid fusion of face and iris is proposed. A novel feature level fusion of face and iris features, using both low and high frequency sub-bands of Discrete Wavelet Transform (DWT), and Principal Component Analysis (PCA) is designed. The redundant data resulted from feature fusion of face and iris is overcome by feature transformation through Linear Discriminant Analysis (LDA). The proposed feature level fusion is tested for face databases (ORL and Yale), and iris databases (CASIA and UBIRIS). The performance of the proposed feature level fusion approach is superior to DWT, PCA and Gabor+PCA based fusion methods by exhibiting highest recognition rate of 97% with low dimensionality. Further, a hybrid fusion of feature level and score level fusion methods is proposed to improve the performance of the multimodal biometric system. In comparison to feature level and score level fusion methods, the hybrid fusion method attains highest recognition rate of 99.6% and least Equal Error Rate (EER) of 0.086 for ORL+CASIA database.
Keywords: Multimodal biometrics; Feature level fusion; Sub-band fusion; DWT; PCA; LDA; Hybrid fusion.
Ear recognition based on discriminant multi-resolution image representation
by Hakim Doghmane, Hocine Bourouba, Kamel Messaoudi, El Bay Bourennane
Abstract: The multi-resolution analysis is more appropriate for extracting information from measured data, because it is generally multi-scale in nature. This paper proposes a new approach for ear representation, based on multi-resolution analysis framework. Such representation relies on significantly Gabor wavelet, Local Phase Quantization (LPQ) descriptor and Spatial Pyramid Histogram (SPH) method. First, to capture the local structure in ear image, the Gabor wavelet function with two scales and four orientations is used. Second, to fully explore the blur invariant property and the texture information in different scales and directions spaces, the LPQ operator is applied on the image responses of Gabor filter to get label LPQ images. Third, the SPH of horizontal decomposition is applied for each of them, to obtain local ear feature descriptors. Next, the obtained histograms are normalized. Then, the global representation of ear image is obtained by concatenating all the local feature descriptors. After that, a discriminant representation of ear image is constructed using whitened linear discriminant analysis. Finally, the K-nearest neighbor classifier is used for identification. Experiments conducted on two ear databases (IIT Delhi-1 and IIT delhi-2); show that the proposed approach provides a significant accuracy improvement compared to the state of the art methods.
Keywords: Ear biometric recognition; scale; local phase quantization; Gabor wavelet; spatial pyramid histogram; multi-resolution analysis.
SUPER RESOLUTION AND RECOGNITION OF UNCONSTRAINED EAR IMAGE
by Anand Deshpande, Prashant Patavardhan
Abstract: In this paper, a framework is proposed to super-resolve low resolution ear images and to recognize these images, without external dataset. This frame uses linear kernel co-variance function based Gaussian Process Regression to super-resolve the ear images. The performance of the proposed framework is evaluated on UERC database by comparing and analyzing the peak signal to noise ratio, structural similarity index matrix and visual information fidelity in pixel domain. The results are compared with the state-of-the-art-algorithms. The results demonstrate that the proposed approach outperforms the state-of-the-art super resolution approaches.
Keywords: Super resolution: ear recognition: Gaussian process regression: PSNR.
A Comparison of Human Brainwaves-Based Biometric Authentication Systems
by Shikah Alsunaidi, Nazar Saqib, Khalid Alissa
Abstract: Several decades ago, attention was directed to biometrics as an alternative to passwords that can be discovered or "Shoulder Surfing" by others. Various authentication methods have been provided that rely on the user's biometrics, such as a fingerprint of a face, iris, voice, and others. Unfortunately, ways were found to imitate these visible fingerprints for using them in penetrations. Therefore, many researchers were interested in studying the possibility of using brainwaves for authentication purposes, as relying on hidden vital features increases the difficulty of breaking and imitation. This paper presents an analytical study of the proposed brainwave-based biometric authentication systems. It provides a comparison of signal acquisition methods for the brainwave-based authentication system. Also, the paper classifies brainwaves using its relevant features. It also presents the phases of the brainwave-based authentication system. Finally, it provides a detailed discussion of several factors that affect the accuracy of the brainwave-based authentication system results, and evaluate the compatibility level of the brainwave with the biometric requirements.
Keywords: Authentication; continuous authentication; biometrics; brainwave; BCI; acquisition methods; intracortical; ECoG; EEG; MEG; fMRI; fNIRS.
Investigating the Accuracy of Free-Text Keystroke Dynamics Authentication in Touchscreen Devices
by Suliman Alsuhibany, Muna Almushyti, Fatimah Alkhudhayr
Abstract: Security on smartphones has become a substantial topic since the amount of security relevant data and information stored on them has been increasing. While the password is commonly used as a traditional approach for user authentication, it suffers from a security-usability trade-off dilemma. Accordingly, the keystroke dynamics authentication approach can provide ease of use to the user as well as robust security. This paper investigates the feasibility of utilizing user-typing behaviors on touchscreen keypads for the authentication process via free-text keystroke dynamics. In constructing the timing vectors, three timing features are used: hold time, flight time, and Di-graph duration. Additionally, Euclidian and Manhattan distances were utilized to find the degree of similarity between the users log-in data and users profile. An Android application was developed and evaluated through an experimental study. Thus, we achieved a very encouraging result, as we were able to show that applying the free-text method indeed influences the accuracy of user authentication on a touchscreen device
Keywords: Authentication; Security; Biometrics; Keystroke Dynamics; Touchscreen.
Two-Phase Palmprint Identification
by Hemantha Kumar Kalluri, Prasad M V N K, Arun Agarwal, Raghavendra Rao Chillarge
Abstract: 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.
Keywords: Biometrics; Clustering; Palmprint Identification; Palmprint Recognition.
Multiscale Neighborhood Based-Tree Binary Pattern: A new feature Descriptor for Face Recognition
by Shekhar Karanwal
Abstract: This research paper proposes the novel Face recognition(FR) descriptor for pose and expression variations so-called Multiscale Neighbourhood Based-Tree Binary Pattern(MNB-TBP). To produce the entire feature size of the MNB-TBP descriptor, 2 local descriptors are introduced. These 2 local descriptors are called as NB-TBP and Mean(Mn)NB-TBP. The methodologies adopted for NB-TBP and MnNB-TBP descriptors are totally novel and are based on tree structure. Specifically by considering the 5x5 pixel window the root node is allocated the centre value and at the max 2 children are assigned to the centre, separately from both the scale pixels(i.e. first 2 neighbourhoods of scale(Z1 and Z2)). Then next 2 neighbourhoods are assigned to the left child node and the further 2 neighbourhoods are assigned to the right child node. This concept continues till all the neighbourhoods acquires their position in the tree structure. Therefore from scale Z1 and Z2 2 tree structures are produced. Then all the respective child nodes are compared with the respective parent node to produce the feature size of the NB-TBP descriptor and mean of the respective child nodes are compared with the respective parent node for producing the feature size of the MnNB-TBP descriptor. The multiscale feature extraction and combination of both the proposed descriptors gives the emergence of the MNB-TBP descriptor. To produce the compact and robust feature for classification PCA is applied further. Finally classification is performed by Support vector machines(SVM) and Nearest neighbour(NN). The 3 challenging databases used for the performance evaluation are Olivetti research laboratory(ORL), Georgia technology(GT) and Faces94. The proposed FR approach achieves very effective results which comprehensively outperforms several of state of the art approaches from the literature.
Keywords: Neighborhood based tree binary pattern; Mean neighborhood based tree binary pattern; Multiscale neighborhood based tree binary pattern; Principal component analysis; Support vector machines and Nearest neighbor.
Proposition of new secure data communication technique based on huffman coding, chaos and lsb
by Abdelkader Bouguessa, Naima Hadj Said, Adda Ali Pacha
Abstract: As the number of Internet users grows, finding robust and secure datarncommunication is becoming increasingly popular today. This issue has led to the creation of hybrid security mechanisms. There are several methods in the literature that use various methods of encryption and steganography with certain advantages and disadvantages. This work provides a new hybrid security mechanism that tries to integrate the theory of chaos as cryptography mechanism, with LSB steganography technique. Huffman coding has also been used to increase the ability to integrate the proposed mechanism. another new think in this work is that we use a specific presentation to send plaintext data inside a picture. Our proposals are tested in MATLAB. To examine the effectiveness of the proposed technology, three types of analysis are performed: security, robustness and efficiency analysis. Modelling and results show that the proposed method is better to other methods in the literature.rn
Keywords: Compression; Cryptography; Steganography; Huffman; Chaos; LSB.
Analysing Muzzle Pattern Images as Biometric for Cattle Identification
by Worapan Kusakunniran, Anuwat Wiratsudakul, Udom Chuachan, Thanandon Imaromkul, Sarattha Kanchanapreechakorn, Noppanut Suksriupatham, Kittikhun Thongkanchorn
Abstract: Identifying individual animals is important for many reasons of population control, illegal trade prevention, and disease surveillance. This paper focuses on the cattle identification, using biometric-based solution of muzzle images. The proposed method begins with localizing muzzle region in each image using the haar-cascade based classifier. The scale-invariant feature transform (SIFT) is applied to extract key points of muzzle patterns. Then, SIFT points are split into different clusters/types of muzzle patterns, called bags of muzzle-words (BoM). Finally, the support vector machine (SVM) model is built on BoM as the cattle identifier. The proposed method is evaluated on the published muzzle images dataset of cattles and the collected muzzle image dataset of slaughterhouses and preserved muzzles of swamp buffalo. This reports the perfect accuracy of 100%. It is also evaluated with the collected dataset of muzzle images of swamp buffalo in the real fields with the reported accuracy of above 90%.
Keywords: Cattle Identification; Muzzle Images; Animal Biometric.
LAHAR-CNN: Human Activity Recognition from one image using Convolutional Neural Network Learning Approach
by Hend Basly, Wael Ouarda, Fatma Ezahra Sayadi, Bouraoui Ouni, Adel M. Alimi
Abstract: The problem of human action recognition has attracted the interest of several researchers due to its significant use in many applications. With the great success of deep learning methods in most areas, researchers decided to switch from traditional methods based hand-crafted feature extractors to recent deep learning-based techniques to recognize the action. In the present research work, we propose a Learning Approach for Human Activity Recognition in the elderly based on Convolutional Neural Network (LAHAR-CNN). The CNN model is used to extract features from the dataset, then, a Multilayer Perceptron (MLP) classifier is used for action classification. It has been widely admitted that features learned using a CNN model on a large dataset can be successfully transferred to an action recognition task with a small training dataset. The proposed method is evaluated on the well-known MSRDailyActivity 3D dataset. It has shown impressive results that exceed the performances obtained in the state of the art using the same dataset, thus reaching 99.4%. Furthermore, our proposed approach predicts human activity (HA) from one single frame sample which justifies its robustness. Hence, the proposed model is ranked at the top of the list of space-time techniques.
Keywords: Human activity recognition; Convolutional Neural Network; Deep learning; Daily Living Activity.
Special Issue on: Biometrics Challenges and Applications
Experimental results on palmvein-based personal recognition by multi-snapshot fusion of textural features
by Mohanad Abukmeil, Gian Luca Marcialis
Abstract: In this paper, we investigate multiple snapshot fusion of textural features for palmvein recognition including identification and verification. Although the literature proposed several approaches for palmvein recognition, the palmvein performance is still affected by identification and verification errors. As well-known, palmveins are usually described by line-based methods which enhance the vein flow. This is claimed to be unique from person to person. However, palmvein images are also characterized by texture that can be pointed out by textural features, which relies on recent and efficient algorithms such as Local Binary Patterns, Local Phase Quantization, Local Tera Pattern, Local directional Pattern, and Binarized Statistical Image Features (LBP, LPQ, LTP, LDP and BSIF, respectively), among others. Finally, they can be easily managed at feature-level fusion, when more than one sample can be acquired for recognition. Therefore, multi-snapshot fusion can be adopted for exploiting these features complementarity. Our goal is to show that thisrnis confirmed for palmvein recognition, thus allowing to achieve very high recognition rates on a well-known benchmark data set.
Keywords: Palmvein Recognitioon; multi-snapshot fusion; local dense descriptor.
FACE DETECTION AND RECOGNITION SYSTEM BASED ON HYBRID STATISTICAL, MACHINE LEARNING AND NATURE BASED COMPUTING
by Vinodini Ramamurthy, Karnan M.
Abstract: Face detection becomes an important task carried out in biometric based security system and identification application. This paper presents the detailed investigations on different methods suffer from accuracy and computational complexity used for the face detection and recognition. The face detection and recognition with high performance ratio for face detection and recognition is achieved in the methods investigated. The reduction of complexity can happen at any stages of the face recognition like preprocessing, segmentation, feature extraction, recognition etc. The proposed method presented in this paper is based on PCA (principle component analysis), SVM (support vector machine), K-nearest neighbor (KNN) and ACO (ant colony optimization). The detail investigation of the proposed method is made and is compared with the existing methods. From the performance it can be observed that the proposed method is better in performance when compared to other methods.
Keywords: Face detection; recognition; PCA; SVM; ACO; segmentation; feature extraction; classification.
Special Issue on: Biometrics, Deep Learning and Sentiment Analysis
Cascading Failure of Complex Networks under Degree-based Attack
by Yan Liu, Jie Yang, Peng Geng
Abstract: Frequent large-scale blackouts, network disconnections, and traffic jams have led to an increasing focus on the cascading failure of complex networks. This paper summarizes the research status of cascading failure in communication networks, power grids and independent networks. Considering the load and capacity of nodes, a cascading failure model of complex networks for degree-based attack is established. In this model, a new measurement method is proposed. This method is a normalized index to measure the impact of cascading failure, which is used to study the global consequences of different types of node failure. By attacking the nodes with higher and lower degree values, the effects of Higher-Degree-Based Attack (attacking nodes with a higher degree in descending order) strategy and Lower-Degree-Based Attack (attacking nodes with a lower degree in ascending order) strategy on network robustness are discussed. The simulations show that when certain conditions are met, attacking nodes with a lower degree (Lower-Degree-Based Attack) can cause greater damage. This result breaks the conventional thinking and provides a reference for the protection of important nodes in the real complex networks.
Keywords: cascading failure; complex networks; degree-based attack.
Special Issue on: Recent Trends in Pattern Recognition and Biometrics
Feature Recognition Method For Similar Key Points Of Human Face Based On Adaptive Median Filter
by Jing Liu
Abstract: In order to overcome the low efficiency and poor accuracy of current face feature recognition methods, this paper proposes a feature recognition method for similar key points of human face based on adaptive median filter. The preprocessing of face image including rotation, scaling and clipping was carried out to remove the influence of image background on feature recognition of similar key points of face. The gray level of the input image is mapped to the output image pixel by function mapping to achieve histogram homogenization of the gray face image.Based on adaptive median filtering, a face image denoising model is constructed and trained.The experimental results show that the recognition time of the proposed method is less than 0.7s, the SNR is higher than 24dB, the recognition accuracy is more than 90%, and the recognition effect is better.
Keywords: Adaptive median filter; Similar key points of human face; Feature recognition; Image denoising.
Research On Fast Identification Technology Of Forged Fingerprints Based On The Improved K-Means Algorithm
by Zhao-ting Ren
Abstract: In order to overcome the low accuracy of the traditional method, a fast identification method based on the improved k-mean algorithm is proposed. Spatial grid block model is constructed to extract the fingerprint texture features and then the fingerprint profile features are detected using the edge outline extraction method. The Kalman fusion method is used to reconstruct fingerprint information. Using the neighborhood distributed retrieval method, fingerprint image feature fusion is realized and the texture feature extraction model for forged fingerprints is established. The K-means clustering method is used for fingerprint feature clustering to realize fast identification of forged fingerprints. Experimental results show that the identification accuracy of this method is higher than 0.85, and the identification stability is good. The signal-to-noise ratio of fingerprint images is always between 25.3dB and 82.3dB, and the imaging quality is high, indicating that this method can realize fast and accurate identification of forged fingerprints.
Keywords: K-means algorithm; forged fingerprint; fast identification; feature extraction; texture.
Method for accurately identifying local fuzzy features of sprinting video images
by Zhiling Chen
Abstract: In order to improve the recognition ability of sprinter video image features, a method of image local fuzzy feature recognition based on edge contour feature matching was designed. Based on the model of image visual feature sampling, the spatial block region planning is carried out. The attitude determination model of the local fuzzy region is established, and the local fuzzy features are extracted by combining template matching and wavelet multi-scale decomposition. Block recognition and information enhancement technology are used to enhance the fuzzy region information so as to extract the edge contour feature set of the fuzzy region and realize the accurate recognition of the local fuzzy features of the image. The simulation results show that this method can accurately identify the local fuzzy features of sprint video images, and the highest recognition accuracy can reach 95.7%.
Keywords: Sprinting; video image; local fuzzy feature; accurate identification; edge contour detection.
Research On Fingerprint Feature Recognition Of Access Control Based On Deep Learning
by Xiaochang Lv, Li Ding, Guohua Zhang
Abstract: In order to overcome the problems of large error and long time-consuming in traditional feature recognition methods, this paper proposes a new fingerprint feature recognition method based on deep learning. Firstly, fingerprint identity database is established, and the access control fingerprint image is collected by the modified hardware equipment, and the image preprocessing is realized from two aspects: image screening and morphological processing. In this framework, the fingerprint direction field in the fingerprint image is screened through multiple iterations. The feature points in the fingerprint image of access control are extracted, and the similarity between the image and the information base is calculated. The experimental results show that compared with the traditional recognition method, the recognition speed of the proposed method is improved by about 6.6 seconds on the premise of ensuring the accuracy of recognition.
Keywords: Deep learning; Access control fingerprint; Fingerprint feature; Feature recognition;.
Research On Fingerprint Image Recognition Based On Convolution Neural Network
by Xin Zheng
Abstract: In order to overcome the problem of poor image matching performance of the image recognition method, a method of fingerprint image recognition based on convolution neural network is proposed. In this method, the defaced fingerprint image is preprocessed by smoothing, convergence, equalization, background foreground segmentation and distortion correction, and the feature points of the defaced fingerprint image are extracted by combining the neighborhood judgment method, and the information pseudo feature points are removed by fusing the feature points, the center points are extracted from the feature points of the defaced fingerprint image, and the center block image is identified by convolution neural network, so as to realize the defaced fingerprint image Distinguish. The experimental results show that the performance of restoration and reconstruction is improved. The rejection rate (FRR) is 3.75%, the false recognition rate (far) is 1.25%, and the correct recognition rate (CR) is 99.25%.
Keywords: Convolution neural network; Defaced fingerprint; Image recognition; Neighborhood determination.
Unconstrained Online Handwritten Uyghur Word Recognition Based On Recurrent Neural Network And Connectionist Temporal Classification
by Mayire Ibrayim, Wujiahemaiti Simayi, Askar Hamdulla
Abstract: This paper conducts the first experiments applying recurrent neural networks-RNN accompanied with Connectionist temporal classification-CTC to build end-to-end online Uyghur handwriting word recognition system. The traced pen-tip trajectory is fed to network without conducting segmentation and feature extraction. The network is trained to transcribe handwritten word trajectory to a string of characters in alphabet which has total 128 character forms. In order to avoid overfitting during training and improve generalization of the model, dropout technique is implemented. An online handwritten word dataset has been established and used for model training and evaluation in writer independent manner. Recognition results are evaluated by calculating the Levenshtein-edit distance and 14.73% character error rate CER on test set of 3600 samples for 900 word classes has been observed without help of any lexicon search and language model.
Keywords: Online Handwriting Recognition; Recurrent Neural Networks; Connectionist Temporal Classification; Dropout; Uyghur Words.
Research On Facial Feature-Based Gender Intelligent Recognition Based On The Adaboost Algorithm
by Jing Wang
Abstract: In order to overcome the problem of poor facial recognition intelligence and weak gender judgment, a new method based on Adaboost algorithm for facial feature-based gender intelligence recognition is proposed in this paper. In this method, the three-dimensional special point detection, weak perspective projection, spatial region segmentation and other methods are employed to construct the facial feature information sampling model. The Adaboost algorithm is used to analyze the matching between facial features and gender, on which facial-feature gender intelligent recognition is performed according to the the distribution of the eyes, nose and mouth of the face image, and the edge contour detection model of the face image is constructed. The experimental results show that the method has the advantages of good intelligence, high recognition precision and short time cost in face-based gender recognition.
Keywords: Adaboost algorithm; facial features; gender; intelligent identification.
Research On Facial Expression Recognition Of Video Stream Based On Opencv
by Feng Gao, Daizhong Luo, Xinqiang Ma
Abstract: In order to overcome the poor performance of expression similarity measurement in traditional video stream facial expression recognition methods, an opencv based facial expression recognition method is proposed. In this method, the video stream face detection image is obtained by the window detection of various features in each position for the video stream image through the cascade classifier, and the image preprocessing is implemented. Based on OpenCV, the most important eyes and mouth in the facial expression are modeled, the eye feature model and mouth feature model are constructed, and the facial expression recognition of the video stream is realized through the constructed model. The experimental results show that the performance of expression similarity measurement is better, and the recognition rate of different expressions is more than 90%.
Keywords: OpenCV; Video stream; Face; Facial expression recognition;.
Research On The Application Of Convolutional-Deep Neural Networks In Parallel Fingerprint Minutiae Matching
by SuHua Wang, MingJun Cheng, ZhiQiang Ma, XiaoXin Sun
Abstract: In order to overcome the problem of low throughput and time-consuming of traditional fingerprint minutiae matching methods, a new convolutional-deep neural network is proposed for parallel fingerprint minutiae matching. This method realizes the preprocessing of the initial image through four steps: normalization, image enhancement, parallel thinning and image segmentation. The convolutional-deep neural network is constructed from convolution kernel, convolution layer, pooling layer and full connection layer to extract minutiae of fingerprint image. Through feature minutiae matching, local matching and global matching, the matching results of fingerprint parallel nodes are obtained. The experimental results show that compared with the traditional matching method, the fingerprint matching throughput of convolutional-deep neural network is increased by 25%, and the matching time is saved by about 8 seconds.
Keywords: Convolution network; Deep neural network; Fingerprint minutiae; Parallel matching; Minutiae matching;.
Special Issue on: Bio-Inspired Algorithms for Biometrics
Face Spoofing Detection Using Improved SegNet Architecture with Blur Estimation Technique
by Sandeep Kumar, Sukhwinder Singh, Jagdish Kumar
Abstract: Biometrics has been increasingly used as the well-known technology for the identification and verification of a person. The huge demand of biometric in day by day life, cybercrime is going increase rapidly in the digital world. Among different biometric traits face has been extensively used for human identity and is therefore much vulnerable to face spoofing attacks. In this spoofing attack, Fake printed photo of user is presented in front of camera. In this proposed work, face liveness detection (FLD) scheme on photo attack using convolution neural network with texture-based blur estimation feature & elimination using Support Vector Machine (SVM). The face is detected with the help of improved SegNet based convolutional neural network (CNN) method. Blur measure on the basis of local min-max of left and right edges and Calculate blur of horizontal and vertical edges. Image filtering is done by adaptive median filter (AMF). The proposed & novel 5-Layer encoder decoder SegNet based algorithm improves the accuracy on various benchmark dataset i.e. NUAA, Replay, Printed, CASIA and live database for face liveness detection. The detection rate has reached up to 97% and time taken for liveness reduced up to one sec per image. This proposed algorithm shows better value of recall, precision and error rate as compared to earlier algorithms
Keywords: Face Liveness; SVM; Blur Estimation; CNN; Adaptive Median Filter; Face Detection.
Image Recognition Method For Fault Service Action Of Tennis Based On Feature Matching
by Shouguo Jin
Abstract: In order to overcome the problems of low accuracy and long time-consuming in the traditional image recognition method for fault service action of tennis, this paper proposes a new image recognition method for fault service action of tennis based on feature matching. This method firstly collects and processes the tennis service image, including image graying, image denoising, background difference, etc., then uses Harris operator to extract and describe the image feature points, and finally realizes the feature matching by measuring the similarity of the image feature points to complete the fault action recognition of tennis service. The experimental results show that compared with the traditional method, the proposed method improves the recognition accuracy and recognition efficiency greatly, and the shortest recognition time is only 10s, which provides a favorable basis for improving tennis players' service technology, and has a wide range of application value.
Keywords: Feature matching; Tennis image; Fault service; Action recognition method.
A gait recognition method for moving target image in sports based on decision tree
by Zhaoxiang Zhang
Abstract: In order to overcome the low accuracy of traditional gait recognition methods, a new gait recognition method based on decision tree is proposed. The decision tree classification method is used to obtain the effective gait features in the sports target image; based on the acquired effective gait features, the weighted block sparse representation method is used to realize the gait recognition of the sports target image. The experimental results show that this method has a high recognition accuracy for the gait in the sports target image with small area occlusion and no occlusion. When the dimension is 301, the recognition accuracy of this method is 97.5%. Compared with the similar recognition methods, the recognition accuracy of this method is significant and can be used in the task of gait recognition of sports target image.
Keywords: Decision tree; Sports; Moving target; Image; Gait; Recognition.
Target Tracking And Recognition Of Moving Video Image Based On Convolution Feature Selection
by JunWei Yang
Abstract: In order to overcome the low accuracy of moving video image target tracking and recognition, a method of moving video image target tracking and recognition based on convolution feature selection is proposed. In this method, feature centers are generated according to the distance matrix between feature images, and feature dimensions are compressed. The multi-layer convolution feature is used to train multiple trackers to jointly determine the target state. The weight of the tracker is updated online by the real-time error of the tracker, and the information redundancy and noise between different convolution features are filtered out. The experimental results show that the recall rate is close to 100% of the success rate of the tracker, the recognition error rate is close to 0, and the recognition time is less than 0.5min, which can effectively improve the recognition accuracy. At the same time, the whole algorithm has strong adaptability.
Keywords: Convolution feature selection; Moving video image; Target tracking and recognition.
Recognition Algorithm Of Athlete's Partially Occluded Face Based On Deep Learning Algorithm
by Wenjuan Li, Kevin Millsap
Abstract: Because there are a lot of noise data in the partially occluded face image, the existing recognition methods have the problems of low recall rate and long time consuming. In this paper, a new recognition algorithm based on depth learning algorithm is proposed. This method uses boosting algorithm to locate face information, based on which the face image is grayed and denoised. The local binary pattern is used to extract face features, and the convolution neural network in deep learning algorithm is used to realize face feature recognition. The experimental results show that compared with the traditional face feature recognition algorithm, the proposed method has significantly improved recognition accuracy and recall rate, and the feature recognition time is shorter, which proves that the proposed algorithm has stronger application performance.
Keywords: Deep learning algorithm; Convolution neural network; Athlete; Local occlusion; Face feature; Recognition algorithm.
Research On Emotion Recognition Method Of Weightlifter Based On Non-Negative Matrix Decomposition Algorithm
by Qiang Wu
Abstract: In order to overcome the problems of high recognition error rate and low recognition efficiency existing in the traditional method of emotional recognition of weightlifters, a method of emotional recognition of weightlifters based on non negative matrix decomposition algorithm is proposed. The database of emotion recognition matching criteria is established, and the mapping relationship among emotion, facial expression of far mobilization and weight lifting posture is formed. The real-time motion images of weightlifters are collected, and the image classifier is designed by using the non negative matrix decomposition algorithm. The real-time images are divided into facial expression images and whole body motion images, and the features of the two images are extracted respectively. The extracted results are matched with the results of the standard database, and the real-time emotion recognition results of weightlifters are obtained. The experimental results show that compared with the traditional emotion recognition method, the proposed method improves the accuracy of emotion recognition by 50%.
Keywords: Non-negative matrix decomposition algorithm; Weightlifter; Athlete emotion; Mapping relationship; Dimension reduction.
An Analysis Of Mandarin Emotional Tendency Recognition Based On Expression Spatiotemporal Feature Recognition
by Caihua Chen
Abstract: In order to overcome the problem of high recognition error rate in traditional emotional tendency recognition methods, a Mandarin emotional tendency recognition method based on expression spatiotemporal feature recognition is proposed. This method extracts the spatiotemporal features of the expression of the research object, and uses the data fusion technology to fuse the extracted feature vector. This paper constructs the standard database of Mandarin emotional tendency recognition, and takes the database as the standard of emotional tendency recognition. The fusion feature vector is matched with the standard feature in the database to get the result of Mandarin emotional tendency recognition. The experimental results show that compared with the traditional method for Mandarin emotional tendency recognition, the recognition method based on the spatiotemporal feature of expression can reduce the recognition error by about 75%.
Keywords: Spatiotemporal features of expression; Data fusion; Cognition; Mandarin; Emotion recognition.
The Method of Table Tennis Players' Posture Recognition Based On Genetic Algorithm
by Jiesen Yin
Abstract: In order to overcome the problems of poor denoising effect, low recognition efficiency and low recognition accuracy of the existing methods of table tennis players' posture recognition, this paper proposes a method of table tennis players' posture recognition based on genetic algorithm. The optimization model of multi-objective key frame extraction is constructed, and the optimization model of multi-objective key frame extraction is solved by genetic algorithm to obtain the key frame. Kalman filter is used to remove the noise in the key frame, eliminate the interference of the noise on the recognition result, and shorten the recognition time. According to the results of noise removal, the dynamic time warping algorithm is used to recognize table tennis players' posture. The experimental results show that the proposed method has good denoising effect, high recognition efficiency and high recognition accuracy, with the highest recognition accuracy of 98.7%.
Keywords: Genetic algorithm; Table tennis; Athletes; Pose recognition; Key frame extraction.
Feature Similarity Measurement Of Cross-Age Face Image Based On Deep Learning Algorithm
by Dong-yuan Ge, Xi-fan Yao, Wen-jiang Xiang, En-chen Liu, Zhi-bin Xu
Abstract: In order to overcome the problem of low accuracy of traditional facial image feature similarity measurement methods, the paper proposes a new cross-age facial image feature similarity measurement method based on deep learning algorithm. Face segmentation is performed based on eye coordinates to determine the effective area for face detection. Introduce deep learning algorithms to build the basic architecture of deep learning networks, define network input as face image data containing age and identity tags in a cross-age database, and define output as face image features under constant age, and pass the foundation separately Training and cross-age training complete the network training, and calculate the facial features under the same age. The cosine distance method is used to measure the similarity of face image features across ages. The experimental results show that the method has obtained more accurate measurement results in different age data sets, and the highest measurement accuracy is 97.9%.
Keywords: Deep learning algorithm; Face image; Similarity measurement; Cosine distance method.
Multi View Face Pose Recognition Model Construction Based On Typical Correlation Analysis Algorithm
by Rongtao Liao, Yuzhe Zhang, Yixi Wang, Dangdang Dai
Abstract: In order to overcome the problems of large recognition error and low recognition accuracy in the existing face pose recognition model, the paper proposes and constructs a new multi-view face pose recognition model based on typical correlation analysis algorithm. First, the AdaBoost algorithm is used to realize multi-pose face detection and positioning. Secondly, the face image is preprocessed, including image graying, image denoising, and face image geometric normalization, and then the typical correlation analysis algorithm is used to extract face features Finally, multi-view facial gesture recognition is realized through convolutional neural network. Experimental results show that, compared with the traditional recognition model, the recognition accuracy of the constructed model is greatly improved, and the average accuracy (mAP) is 96.334%, which proves that the recognition performance of the constructed model is better.
Keywords: canonical correlation analysis algorithm; convolution neural network; multi view face; gesture recognition;.
Feature Extraction Method Of Face Image Texture Spectrum Based On Deep Learning Algorithm
by Suhua Wang, Zhiqiang Ma, Xiaoxin Sun
Abstract: Deep learning has made great progress in the field of face recognition, but most of the current face feature matching algorithms focus on the matching of single image and single image, and can not effectively use the relevant information between image sequences, in order to avoid the influence of human factors on the skin texture feature extraction of face image. In this paper, a texture spectrum feature extraction method based on deep learning is proposed. The face image is extracted by CNN network, and the similar image sequences are automatically selected for feature matching by using the improved sparse expression method to obtain the relevant information between the face image sequences. The experimental results show that the algorithm has achieved good results in LFW and AR databases and is superior to the traditional SRC, L1 norm and crc-rls algorithms.
Keywords: Face Image Texture Spectrum Feature; Constrained Sparse Representation; Deep Learning; Image Sequence.