Traditional Hand-crafted Features Research Articles

Automated embolic signal detection using Deep Convolutional Neural Network.

This work investigated the potential of Deep Neural Network in detection of cerebral embolic signal (ES) from transcranial Doppler ultrasound (TCD). The resulting system is aimed to couple with TCD devices in diagnosing a risk of stroke in real-time with high accuracy. The Adaptive Gain Control (AGC) approach developed in our previous study is employed to capture suspected ESs in real-time. By using spectrograms of the same TCD signal dataset as that of our previous work as inputs and the same experimental setup, Deep Convolutional Neural Network (CNN), which can learn features while training, was investigated for its ability to bypass the traditional handcrafted feature extraction and selection process. Extracted feature vectors from the suspected ESs are later determined whether they are of an ES, artifact (AF) or normal (NR) interval. The effectiveness of the developed system was evaluated over 19 subjects going under procedures generating emboli. The CNN-based system could achieve in average of 83.0% sensitivity, 80.1% specificity, and 81.4% accuracy, with considerably much less time consumption in development. The certainly growing set of training samples and computational resources will contribute to high performance. Besides having potential use in various clinical ES monitoring settings, continuation of this promising study will benefit developments of wearable applications by leveraging learnable features to serve demographic differentials.

Device-Free Wireless Localization and Activity Recognition: A Deep Learning Approach

Device-free wireless localization and activity recognition (DFLAR) is a new technique, which could estimate the location and activity of a target by analyzing its shadowing effect on surrounding wireless links. This technique neither requires the target to be equipped with any device nor involves privacy concerns, which makes it an attractive and promising technique for many emerging smart applications. The key question of DFLAR is how to characterize the influence of the target on wireless signals. Existing work generally utilizes statistical features extracted from wireless signals, such as mean and variance in the time domain and energy as well as entropy in the frequency domain, to characterize the influence of the target. However, a feature suitable for distinguishing some activities or gestures may perform poorly when it is used to recognize other activities or gestures. Therefore, one has to manually design handcraft features for a specific application. Inspired by its excellent performance in extracting universal and discriminative features, in this paper, we propose a deep learning approach for realizing DFLAR. Specifically, we design a sparse autoencoder network to automatically learn discriminative features from the wireless signals and merge the learned features into a softmax-regression-based machine learning framework to realize location, activity, and gesture recognition simultaneously. Extensive experiments performed in a clutter indoor laboratory and an apartment with eight wireless nodes demonstrate that the DFLAR system using the learned features could achieve 0.85 or higher accuracy, which is better than the systems utilizing traditional handcraft features.

Learning cast shadow appearance for human posture recognition

This paper presents a system for human posture recognition using a camera and two infrared light sources. It uses as input the combination of the body silhouette and its (invisible to the eye) cast shadows. Conventional video-surveillance methods based on a single camera can fail to infer the correct posture since different postures can look similar under perspective projection. Fortunately, cast body shadows, generated by infrared lights, offer additional posture information that cannot be directly captured by a single camera. Each shadow can be projected on different surfaces (e.g. floor, walls, and furniture) generating complex body projections that represent various shapes within the same posture class. These shadow images are very challenging and difficult to describe with traditional handcrafted features that need to be somewhat invariant to these within-class changes. However, a deep convolution neural network (CNN) is able to learn a better data representation from a large-scale dataset. In the absence of a big real dataset, we propose to use synthetic data for training the CNN classifier. Learning from synthetic data is a challenging task due to the gap between synthetic and real feature distributions. Thus, we propose a normalization technique to bridge the gap and help the classifier to better generalize with real data. We evaluated the proposed system on a new real dataset captured in our laboratory and a simulated dataset generated with computer graphics tools. Experimental results validated the efficiency of the CNN model against other conventional methods. Furthermore, the combination of cast shadows and body silhouette had better performance than using only the body silhouette as expected.

Physiognomy: Personality traits prediction by learning

Evaluating individuals’ personality traits and intelligence from their faces plays a crucial role in interpersonal relationship and important social events such as elections and court sentences. To assess the possible correlations between personality traits (also measured intelligence) and face images, we first construct a dataset consisting of face photographs, personality measurements, and intelligence measurements. Then, we build an end-to-end convolutional neural network for prediction of personality traits and intelligence to investigate whether self-reported personality traits and intelligence can be predicted reliably from a face image. To our knowledge, it is the first work where deep learning is applied to this problem. Experimental results show the following three points: 1) “Rule-consciousness” and “Tension” can be reliably predicted from face images. 2) It is difficult, if not impossible, to predict intelligence from face images, a finding in accord with previous studies. 3) Convolutional neural network (CNN) features outperform traditional handcrafted features in predicting traits.

Handcrafted vs. non-handcrafted features for computer vision classification

This work presents a generic computer vision system designed for exploiting trained deep Convolutional Neural Networks (CNN) as a generic feature extractor and mixing these features with more traditional hand-crafted features. Such a system is a single structure that can be used for synthesizing a large number of different image classification tasks. Three substructures are proposed for creating the generic computer vision system starting from handcrafted and non-handcrafter features: i)one that remaps the output layer of a trained CNN to classify a different problem using an SVM; ii) a second for exploiting the output of the penultimate layer of a trained CNN as a feature vector to feed an SVM; and iii) a third for merging the output of some deep layers, applying a dimensionality reduction method, and using these features as the input to an SVM. The application of feature transform techniques to reduce the dimensionality of feature sets coming from the deep layers represents one of the main contributions of this paper. Three approaches are used for the non-handcrafted features: deep transfer learning features based on convolutional neural networks (CNN), principal component analysis network (PCAN), and the compact binary descriptor (CBD). For the handcrafted features, a wide variety of state-of-the-art algorithms are considered: Local Ternary Patterns, Local Phase Quantization, Rotation Invariant Co-occurrence Local Binary Patterns, Completed Local Binary Patterns, Rotated local binary pattern image, Globally Rotation Invariant Multi-scale Co-occurrence Local Binary Pattern, and several others. The computer vision system based on the proposed approach was tested on many different datasets, demonstrating the generalizability of the proposed approach thanks to the strong performance recorded. The Wilcoxon signed rank test is used to compare the different methods; moreover, the independence of the different methods is studied using the Q-statistic. To facilitate replication of our experiments, the MATLAB source code will be available at (https://www.dropbox.com/s/bguw035yrqz0pwp/ElencoCode.docx?dl=0).

Urdu Nastaliq recognition using convolutional–recursive deep learning

Recent developments in recognition of cursive scripts rely on implicit feature extraction methods that provide better results as compared to traditional hand-crafted feature extraction approaches. We present a hybrid approach based on explicit feature extraction by combining convolutional and recursive neural networks for feature learning and classification of cursive Urdu Nastaliq script. The first layer extracts low-level translational invariant features using Convolutional Neural Networks (CNN) which are then forwarded to Multi-dimensional Long Short-Term Memory Neural Networks (MDLSTM) for contextual feature extraction and learning. Experiments are carried out on the publicly available Urdu Printed Text-line Image (UPTI) dataset using the proposed hierarchical combination of CNN and MDLSTM. A recognition rate of up to 98.12% for 44-classes is achieved outperforming the state-of-the-art results on the UPTI dataset.

Image-based dementia disease diagnosis via deep low-resource pair-wise learning

Medical Images data is widely acknowledged for its large volumes, high complexities and limited labels information, which makes learning-based tasks challenging to obtain promising performance based on such images data. Also, Alzheimer’s disease, one of the five most severe non-communicable diseases all around the world, receives much attention in both academic research and clinical diagnosis for the time being. It is generally acknowledged that magnetic resonance imaging, a popular imaging tool because of its free of radiation exposure in terms of patients safety issues, has been vastly employed in Alzheimer’s disease diagnosis in recent years. However, most contemporary clinical diagnosis efforts still heavily rely on clinicians’ expertise to realize the disease diagnosis task. In this study, inspired by the recent rapid development of deep learning and low-resource learning techniques, a novel image-based disease diagnosis scheme is proposed to automatically fulfil the diagnosis of Alzheimer’s disease based on arterial spin labeling magnetic resonance images. This new end-to-end low-resource learning scheme is conducted following four steps. First, raw arterial spin labeling images need to be acquired and essential pre-processing steps should be carried out on raw images. Second, the popular convolutional neural network technique is adopted to automatically extract latent trainable features from processed arterial spin labelling images, instead of traditional hand-crafted features. Third, a novel low-resource pair-wise learning method inspired by the semi-supervised learning theory is carried out based on latent trainable features for automatically determining disease similarity functions. Forth, learned outcomes are then fed into a multi-nominal logistic regression model and a regression process with the highest posterior probability is capable to reveal the corresponding disease diagnosis outcome. Extensive experiments based on a database composed of magnetic resonance images acquired from 350 real demented patients are carried out with the newly poroposed scheme being compared towards several other well-known diagnosis tools. All diagnosis results have undergone rigorous and comprehensive statistical analysis composed of analysis of variance and multiple comparison tests. The superiority of the new scheme has been demonstrated therein.

Unsupervised-Learning-Based Feature-Level Fusion Method for Mura Defect Recognition

Mura defect recognition has long been a challenging task in displays, such as the liquid-crystal display (LCD), organic light-emitting diode display and polymer light-emitting diode display. In this paper, we propose an unsupervised-learning-based feature-level fusion approach for mura defect recognition. The approach is known as a joint-feature-representation-based defect recognition framework method. This method concentrates on obtaining effective and sufficient features for mura defects by fusing handcrafted and unsupervised-learned features in a complementary manner. To demonstrate the performance, several experiments are carried out to compare this method with some widely used feature extraction approaches. Experimental results show that the proposed method is more robust and accurate. They also indicate that it is compatible with different unsupervised-learning-based algorithms and handcrafted feature descriptors. Finally, the proposed method is implemented in the vision inspection equipment for recognizing mura defects in thin-film-transistor-LCD panels. It exhibits high robustness and improves the recognition performance by nearly 20% compared with the traditional handcrafted feature descriptors.

Distracted Driver Detection: Deep Learning vs Handcrafted Features

According to the National Highway Traffic Safety Administration, one in ten fatal crashes and two in ten injury crashes were reported as distracted driver accidents in the United State during 2014. In an attempt to mitigate these alarming statistics, this paper explores using a dashboard camera along with computer vision and machine learning to automatically detect distracted drivers. We consider a dataset that incorporates drivers engaging in seven different distracting behaviors using left and/or right hands. Traditional handcrafted features paired with a Support Vector Machine classifier are contrasted with deep Convolutional Neural Networks. The traditional features include a blend of Histogram of Oriented Gradients and Scale-Invariant Feature Transform descriptors used to create Bags of Words. The deep convolutional methods use transfer learning on AlexNet, VGG-16, and ResNet-152. The results yield 85% accuracy with ResNet and 82.5% accuracy with VGG-16, which outperformed AlexNet by almost 10%. Replacing the fully connected layers by a Support Vector Machine classifier did not improve the classification accuracy. The traditional features yielded much lower accuracy than the deep convolutional networks.

Disambiguating named entities with deep supervised learning via crowd labels

Named entity disambiguation (NED) is the task of linking mentions of ambiguous entities to their referenced entities in a knowledge base such as Wikipedia. We propose an approach to effectively disentangle the discriminative features in the manner of collaborative utilization of collective wisdom (via human-labeled crowd labels) and deep learning (via human-generated data) for the NED task. In particular, we devise a crowd model to elicit the underlying features (crowd features) from crowd labels that indicate a matching candidate for each mention, and then use the crowd features to fine-tune a dynamic convolutional neural network (DCNN). The learned DCNN is employed to obtain deep crowd features to enhance traditional hand-crafted features for the NED task. The proposed method substantially benefits from the utilization of crowd knowledge (via crowd labels) into a generic deep learning for the NED task. Experimental analysis demonstrates that the proposed approach is superior to the traditional hand-crafted features when enough crowd labels are gathered.

Learning discriminative representations for semantical crossmodal retrieval

Heterogeneous gap among different modalities emerges as one of the critical issues in multimedia retrieval areas. Unlike traditional unimodal cases, where raw features are extracted and directly measured, the heterogeneous nature of crossmodal tasks requires the intrinsic semantic representation to be compared in a unified framework. Based on a flexible “feature up-lifting and down projecting” mechanism, this paper studies the learning of crossmodal semantic features that can be retrieved across different modalities. Two effective methods are proposed to mine semantic correlations. One is for traditional handcrafted features, and the other is based on deep neural network. We treat them respectively as normal and deep version of our proposed shared discriminative semantic representation learning (SDSRL) framework. We evaluate both of these two methods on two public multimodal datasets for crossmodal and unimodal retrieval tasks. The experimental results demonstrate that our proposed methods outperform the compared baselines and achieve state-of-the-art performance in most scenarios.

Matching-range-constrained real-time loop closure detection with CNNs features.

The loop closure detection (LCD) is an essential part of visual simultaneous localization and mapping systems (SLAM). LCD is capable of identifying and compensating the accumulation drift of localization algorithms to produce an consistent map if the loops are checked correctly. Deep convolutional neural networks (CNNs) have outperformed state-of-the-art solutions that use traditional hand-crafted features in many computer vision and pattern recognition applications. After the great success of CNNs, there has been much interest in applying CNNs features to robotic fields such as visual LCD. Some researchers focus on using a pre-trained CNNs model as a method of generating an image representation appropriate for visual loop closure detection in SLAM. However, there are many fundamental differences and challenges involved in character between simple computer vision applications and robotic applications. Firstly, the adjacent images in the dataset of loop closure detection might have more resemblance than the images that form the loop closure. Secondly, real-time performance is one of the most critical demands for robots. In this paper, we focus on making use of the feature generated by CNNs layers to implement LCD in real environment. In order to address the above challenges, we explicitly provide a value to limit the matching range of images to solve the first problem; meanwhile we get better results than state-of-the-art methods and improve the real-time performance using an efficient feature compression method.

Learning Multilayer Channel Features for Pedestrian Detection.

Pedestrian detection based on the combination of convolutional neural network (CNN) and traditional handcrafted features (i.e., HOG+LUV) has achieved great success. In general, HOG+LUV are used to generate the candidate proposals and then CNN classifies these proposals. Despite its success, there is still room for improvement. For example, CNN classifies these proposals by the fully connected layer features, while proposal scores and the features in the inner-layers of CNN are ignored. In this paper, we propose a unifying framework called multi-layer channel features (MCF) to overcome the drawback. It first integrates HOG+LUV with each layer of CNN into a multi-layer image channels. Based on the multi-layer image channels, a multi-stage cascade AdaBoost is then learned. The weak classifiers in each stage of the multi-stage cascade are learned from the image channels of corresponding layer. Experiments on Caltech data set, INRIA data set, ETH data set, TUD-Brussels data set, and KITTI data set are conducted. With more abundant features, an MCF achieves the state of the art on Caltech pedestrian data set (i.e., 10.40% miss rate). Using new and accurate annotations, an MCF achieves 7.98% miss rate. As many non-pedestrian detection windows can be quickly rejected by the first few stages, it accelerates detection speed by 1.43 times. By eliminating the highly overlapped detection windows with lower scores after the first stage, it is 4.07 times faster than negligible performance loss.