Biometric identification is the technology that differentiates individuals by body parts or behavioral characteristics. Hand has been proved to be a successful biometric for verification and identification because of ...
详细信息
ISBN:
(纸本)9781450389532
Biometric identification is the technology that differentiates individuals by body parts or behavioral characteristics. Hand has been proved to be a successful biometric for verification and identification because of the rich features such as fingerprint, palmprint, dorsal vein, etc. This paper presents a system for identifying individuals based on their hand images. Firstly, after image preprocessing with guided filter and CLAHE method, hand images taken under visible light and near-infrared (NIR) light were normalized. Secondly, a convolutional neural network structure was designed and trained on a large dataset. Using hand images as the input of the network, different depth features were extracted, including the feature from the fusion layer. Thirdly, SVM classifiers were adopted to get the classification results. A fusion strategy was used to make use of different SVM classifiers. The proposed algorithm was tested on different datasets and the experimental results showed that high accuracy can be obtained from the fusion of features. It shows that the hand image is a strong biometric for verification and identification.
The present work describes the use of noninvasive diffuse optical tomography (DOT) technology to measure hemodynamic changes, providing relevant information which helps to understand the basis of neurophysiology in th...
详细信息
The present work describes the use of noninvasive diffuse optical tomography (DOT) technology to measure hemodynamic changes, providing relevant information which helps to understand the basis of neurophysiology in the human brain. Advantages such as portability, direct measurements of hemoglobin state, temporal resolution, non‐restricted movements as occurs in magnetic resonance imaging (MRI) devices mean that DOT technology can be used in research and clinical fields. In this review we covered the neurophysiology, physical principles underlying optical imaging during tissue‐light interactions, and technology commonly used during the construction of a DOT device including the source‐detector requirements to improve the image quality. DOT provides 3D cerebral activation images due to complex mathematical models which describe the light propagation inside the tissue head. Moreover, we describe briefly the use of Bayesian methods for raw DOT data filtering as an alternative to linear filters widely used in signalprocessing, avoiding common problems such as the filter selection or a false interpretation of the results which is sometimes due to the interference of background physiological noise with neural activity.
Colposcopic diagnosis and directed biopsy is the foundation of cervical cancer screening. In the procedure of colposcopy, automatic segmentation of cervical lesion in colposcopic images can provide great assistance an...
详细信息
Convolutional neural networks (CNNs) have proven to be promising in various applications such as audio recognition, image classification, and video understanding. Different dimensions of CNNs (e.g., 1D, 2D, and 3D CNN...
详细信息
ISBN:
(纸本)9781665417488
Convolutional neural networks (CNNs) have proven to be promising in various applications such as audio recognition, image classification, and video understanding. Different dimensions of CNNs (e.g., 1D, 2D, and 3D CNNs) are proposed to adapt to these applications. To accelerate different dimensional convolution, a uniform accelerator is necessary. Nevertheless, the implementation poses a significant challenge due to several observations. Firstly, computational complexity, network mapping methods, and data reuse strategies vary greatly among different dimensional convolutional neural networks. Secondly, various efficient algorithms such as Winograd have been proposed to accelerate CNNs, but their implementations lack flexible support for different network types. Typically, the Winograd-base accelerator is designed for 1-stride and the non-1-stride methods haven’t been implemented on 3D CNNs. To address these challenges, we propose a flexible Winograd-based uniform accelerator (FWUA) for 1D/2D/3D CNNs. With adaptive support for different dimensions, strides, and filter sizes, FWUA is runtime-reconfigurable for different dimensions of CNNs applications, i.e., audio, image, and video. The FWUA is verified on the Xilinx ZCU102 evaluation board FPGA. Our design achieves 1.51/1.13/0.66 (GOPS/DSP) DSP-efficiency and 242/181/105 (GOPS/W) energy-efficiency in C3D, VGG-16, and HAR-CNNs, which are up to 2x comparing to state-of-the-art FPGA works.
This paper proposes an algorithm based on convolutional neural networks for the estimation of the quality level of voice signals transmitted through cellular communication systems. The objective is to take advantage o...
详细信息
ISBN:
(纸本)9781728114910
This paper proposes an algorithm based on convolutional neural networks for the estimation of the quality level of voice signals transmitted through cellular communication systems. The objective is to take advantage of artificial intelligence methods to estimate the MOS parameter and obtain a similar accuracy to that obtained by methods and procedures established in the international norms and international licensed standards. The proposed algorithm uses the MOS results obtained by the method detailed in the ITU-T P.862 standard. The values were obtained for different signals acquired at different reception points. With this information we proceeded to design and train a convolutional neuronal network of 4 layers, achieving very satisfactory results. For the validation, the mean square error was used to measure the degree of similarity of the MOS values obtained by ITU-T P.862 and by the proposed algorithm. The results show a mean square error of 0.00007 for the proposed algorithm.
Edge detection based on ghost imaging technology can directly capture the edge details of a target without acquiring the entire image of the object. In this paper, we propose a method of ghost edge imaging based on un...
详细信息
Existing neural networks for computer vision tasks are vulnerable to adversarial attacks: adding imperceptible perturbations to the input images can fool these models to make a false prediction on an image that was co...
详细信息
Understanding the decision of AI classifiers is fundamental to the reliable and robust application of ML methods across a wide variety of domains and end-uses. This report describes work on a specific area of interest...
详细信息
Understanding the decision of AI classifiers is fundamental to the reliable and robust application of ML methods across a wide variety of domains and end-uses. This report describes work on a specific area of interest conducted under the CMU XAI program, that of detecting and understanding the ability of adversaries to intentionally poison pre-trained classifiers with malicious triggers that allow them full control over the practical use of such systems. We show that by exploiting our developed XAI techniques, it is possible to reliably detect and avoid the use of such classifiers, or indeed to create triggers that are equally capable of breaking the systems. In addition, we present a broader survey of several different approaches to XAI methods, well beyond the scope of the classifier poisoning work, which was additionally developed throughout the course of the program.
Capturing the comprehensive information of various sizes and shapes of images in the same convolution layer is typically a challenging task in computer vision. There are two main kinds of methods for capturing those f...
详细信息
Capturing the comprehensive information of various sizes and shapes of images in the same convolution layer is typically a challenging task in computer vision. There are two main kinds of methods for capturing those features. The first uses the inception structure and its variants. The second utilizes larger convolution kernels on specific layers or stacks with more convolution blocks. However, these methods can result in computationally intensive or vanishing gradients. In this paper, to accommodate feature distributions with different sizes, shapes and reduce computational cost, we propose a width-and depth-aware module named the WD-module to match feature distributions. Moreover, the proposed WD-module consumes less computational cost and parameters compared with traditional residual convolution layers. To verify the effectiveness of our proposed method, a size-and shape-aware backbone network named S(2)A-Net was built, which was obtained by stacking the WD-modules. By visualizing heat maps and features, the proposed S(2)A-Net can adapt to objects with different sizes and shapes in visual recognition tasks and learn more comprehensive characteristics. Experimental results show that the proposed method has higher accuracy in image recognition and outperforms other state-of-the-art networks with the same numbers of layers.
暂无评论