Affective computing is an interdisciplinary research area that includes machine learning and pattern recognition, psychology, and cognitive science. The aim is to research and develop theories, methods and systems that can recognize, interpret, process and simulate human emotions. In this article we propose a neural network model for multimodal emotion recognition based on cross-media data-feature fusion. Multimodal data fusion can effectively improve the accuracy of emotion recognition. We extract features from EEG data and facial images using a deep double-stream neural network and then merge them in a medium-term feature layer to identify three categories of emotions (sadness, calm, and happiness). The experimental results show that the detection accuracy can reach over 95%. Compared to the traditional single-modal emotion recognition method, the accuracy rate of emotion recognition based on EEG data and facial images has been significantly improved. It also proves that the multimodal medium-term feature layer fusion method has good applicability for emotion recognition.
Tongue diagnosis is an important part of the Traditional Chinese Medicine (TCM) diagnosis. In Chinese Medicine, tongue body reflects the most sensitive indicators of the physiological function and pathological changes, which has important application value in the process of diagnosis and treatment of the TCM diagnosis. The accurate separation of tongue body from tongue image is the premise of recognition and diagnosis. Most of the proposed tongue segmentation algorithms are based on the improvement of traditional approaches. These algorithms can improve the segmentation accuracy of tongue image to some extent, but they are less robust. To address above problems, a method of fast tongue image segmentation algorithm using convolutional neural network is proposed in this paper. The network is inspired from ShuffleNet which provided an efficient classification and detection network. The running time of our structure is about 0.16s, and the average segmentation precision is about 90.5%, which makes it of great potential for real-time applications. As opposed to the two common traditional segmentation methods (Kmeans++, GrabCut), the proposed method performs better than the above algorithms.
Tongue diagnosis is an extremely effective and non-invasive auxiliary diagnostic technique. Tongue segmentation is a vital procedure of computer-aided tongue diagnosis system, and its accuracy will directly influence the results of tongue diagnosis. By revising the edge indicator function, in this paper, an improved distance preserving level set evolution (DPLSE) method is proposed for tongue image segmentation. And before segmentation, the grey-level integral projection algorithm, Otus method as well as morphological processing will be utilized to handle tongue images, which to obtain the initial contour curve of this method. The improved algorithm has been experimented on 20 tongue images. Compared with other methods, it shows the competitive performances on tongue images segmentation.
In this paper, we propose a algorithm to tracking target using point feature. The point feature is extracted from the pixels in the first frame and used to label the pixels in the next frame as belonging to either target or background. The positive and negative samples are extracted from the pixels of target and surrounding background, and used to train several weak classifiers, which combine to build a strong classifier using AdaBoost algorithm. The negative samples are given the greater weights than positive samples, which is to avoid that a large number of pixels in background are labeled incorrectly. To efficiently learn a large number of samples, the adopted weak classifier is a linear perceptron model, which is trained and updated using stochastic gradient descent. Only the dot-product between matrices and the sum of matrix elements need to be calculated. To distinguish the similar targets, the histogram-based mean shift algorithm is applied to eliminate those wrong image patches. The histogram of target will be updated over the time. The experiment results show that the proposed algorithm can estimate scale better when scale change, posture change and occlusion occurs.
The traditional algorithms of image super-resolution reconstruction are not effective enough to be used in reconstructing high-frequency information of an image. In order to improve the quality of image reconstruction and restore more high-frequency information, the residual dictionary is introduced which can capture the high-frequency information of images such as the edges, angles and corners. The common dictionary is generated by training and learning pairs of low-resolution and high-resolution images. The dictionary combined by common dictionary and residual dictionary is obtained in which more high-frequency information of the images can be restored while the spatial structure of images can be preserved well. The processing of training and testing dictionary is conducted by Support Vector Regression (SVR). Compared with other algorithms in experiments, the proposed method improves its PSNR and SSIM by 3% ~ 4% and 2% ~ 3% on some different images respectively.
It is necessary to improve the quality of the captured hand vein image in the vein display device and vein recognition system. In this paper, a method of hand vein image enhancement based on phase congruency is proposed according to the structure and features of human hand vein images. Firstly, multiple images containing vein edges are acquired by applying phase congruency which parameters are set differently, and two images that contain the majority of vein and less noise are selected by image entropy values, then the chosen images will be enhanced by contrast enhancement. Finally, the original image and the enhanced image will be fused in gradient domain. The experiment results show that the proposed algorithm can enhance the contrast of the hand vein images efficiently, improve the quality of image significantly, and suppress noise perfectly.
Due to the limitations of image capture device and imaging environments in traditional imaging process, high-resolution (HR) images are difficult to be obtained. The method of digital image processing can be used in image super-resolution with one or an image sequence in original conditions to reconstruct HR images which over the range of imaging system. Traditional learning-based super-resolution algorithm need to run through the sample library with a high computing complexity, and a high recognition rate in the scene with small shifts. This dissertation is mainly about color image SR and parallel implementation of the SR algorithm. An algorithm based on SVM classified learning is proposed in this paper.
Focusing properties of the vector vortex-bearing beams are investigated theoretically by vector diffraction theory. Simulation results show that the intensity distribution in focal region can be altered considerably by adjusting topological charge m and the numerical aperture(NA) of the focusing optical system. Focal pattern evolves from one focal spot to two overlapped intensity peaks1.The two overlapped intensity peaks separate with increasing topological charge m, which leads to the focal splitting. And focal split appears in focal evolution show that the topological charge of the vector vortex-bearing beams influences the focal intensity distribution considerably, and some novel focal patterns appear.
Kinds of factors such as illumination and hand gestures would reduce the accuracy of dorsal hand vein recognition. Aiming at single hand vein image with low contrast and simple structure, an algorithm combining Gabor multi-orientation features fusion with Multi-scale Histogram of Oriented Gradient (MS-HOG) is proposed in this paper. With this method, more features will be extracted to improve the recognition accuracy. Firstly, diagrams of multi-scale and multi-orientation are acquired using Gabor transformation, then the Gabor features of the same scale and multi-orientation will be fused, and the features of the correspondent fusion diagrams will be extracted with a HOG operator of a certain scale. Finally the multi-scale cascaded histograms will be obtained for hand vein recognition. The experimental results show that our method not only improve the recognition accuracy but has good robustness in dorsal hand vein recognition.
In recent years, we have witnessed the prosperity of the face image super-resolution (SR) reconstruction, especially the learning-based technology. In this paper, a novel super-resolution face reconstruction framework based on support vector regression (SVR) about a single image is presented. Given some input data, SVR can precisely predict output class labels. We regard the SR problem as the estimation of pixel labels in its high resolution version. It’s effective to put local binary pattern (LBP) codes and partial pixels into input vectors during training models in our work, and models are learnt from a set of high and low resolution face image. By optimizing vector pairs which are used for learning model, the final reconstructed results were advanced. Especially to deserve to be mentioned, we can get more high frequency information by exploiting the cyclical scan actions in the process of both training and prediction. A large number of experimental data and visual observation have shown that our method outperforms bicubic interpolation and some stateof- the-art super-resolution algorithms.
Parallel processing is the forefront of femtosecond laser micro-nano processing. The key to parallel processing is obtaining multichannel parallel femtosecond laser beams. A method of spatial parallel pulse splitting based on birefringence properties of polarizing splitting prism is proposed for obtaining multichannel parallel ultra-short pulse trains. The generated sub-pulses have the characteristics of equal energy and high similarity. More than that, the compact structure of the polarizing splitting prism makes it easier to be implemented. The accurate relationship between the space interval of pulse sequences and the structural angle, dimension and the distance between the two prisms is mathematically derived. The realizable array form of sub-pulse sequences is theoretically analyzed. The feasibility of the proposed method of femtosecond laser parallel processing is analyzed by software simulation and numerical calculation. The results will provide a new research direction for application of ultrashort pulse in parallel processing.
Hand vein image has been widely used in biological recognition, auxiliary medical and other fields. People with age, height, weight, gender differences have distinction in fat thickness of the back of hand, so the contrast and sharpness of their hand vein images are different too, which may affect the results of applications. In this paper, a hand vein image acquisition system is given and the hand vein images of people from the age of 3 to 60 are obtained in various conditions. The effect on the images caused by ages, genders, BMI (body mass index) and FMI (fat mass index) are researched and the statistical characteristics of the images are analyzed. The types of applicable people are also proposed for applications.
KEYWORDS: Veins, Feature extraction, Canonical correlation analysis, Image fusion, Principal component analysis, Biometrics, Databases, Information fusion, Information technology, Simulation of CCA and DLA aggregates
In this paper, a method of recognition of multi-modal biometrics for palmprint and hand vein based on the feature layer fusion is proposed, combined with the characteristics of an improved canonical correlation analysis (CCA) and two dimensional principal component analysis (2DPCA). After pretreatment respectively, feature vectors of palmprint and hand vein images are extracted using two dimensional principal component analysis (2DPCA),then fused in the feature level using the improved canonical correlation analysis(CCA), so identification can be done by a adjacent classifier finally. Using this method, two biometric information can be fused and the redundancy of information between features can effectively eliminated, the problem of the high-dimensional and small sample size can be overcome too. Simulation experimental results show that the proposed method in this paper can effectively improve the recognition rate of identification.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.