Biblio

Child facial expression recognition is a relatively less investigated area within affective computing. Children’s facial expressions differ significantly from adults; thus, it is necessary to develop emotion recognition frameworks that are more objective, descriptive and specific to this target user group. In this paper we propose the first approach that (i) constructs video-level heterogeneous graph representation for facial expression recognition in children, and (ii) predicts children’s facial expressions using the automatically detected Action Units (AUs). To this aim, we construct three separate length-independent representations, namely, statistical, spectral and graph at video-level for detailed multi-level facial behaviour decoding (AU activation status, AU temporal dynamics and spatio-temporal AU activation patterns, respectively). Our experimental results on the LIRIS Children Spontaneous Facial Expression Video Database demonstrate that combining these three feature representations provides the highest accuracy for expression recognition in children.

At present, pavement crack detection mainly depends on manual survey and semi-automatic detection. In the process of damage detection, it will inevitably be subject to the subjective influence of inspectors and require a lot of identification time. Therefore, this paper proposes the research and implementation of artificial intelligence real-time recognition method of crack edge based on zynq, which combines edge calculation technology with deep learning, The improved ipd-yolo target detection network is deployed on the zynq zu2cg edge computing development platform. The mobilenetv3 feature extraction network is used to replace the cspdarknet53 feature extraction network in yolov4, and the deep separable convolution is used to replace the conventional convolution. Combined with the advantages of the deep neural network in the cloud and edge computing, the rock fracture detection oriented to the edge computing scene is realized. The experimental results show that the accuracy of the network on the PID data set The recall rate and F1 score have been improved to better meet the requirements of real-time identification of rock fractures.

Smart phones have become the preferred way for Chinese Internet users currently. The mobile phone traffic is large from the operating system. These traffic is mainly generated by the services. In the context of the universal encryption of the traffic, classification identification of mobile encryption services can effectively reduce the difficulty of analytical difficulty due to mobile terminals and operating system diversity, and can more accurately identify user access targets, and then enhance service quality and network security management. The existing mobile encryption service classification methods have two shortcomings in feature selection: First, the DL model is used as a black box, and the features of large dimensions are not distinguished as input of classification model, which resulting in sharp increase in calculation complexity, and the actual application is limited. Second, the existing feature selection method is insufficient to use the time and space associated information of traffic, resulting in less robustness and low accuracy of the classification. In this paper, we propose a feature enhancement method based on adjacent flow contextual features and evaluate the Apple encryption service traffic collected from the real world. Based on 5 DL classification models, the refined classification accuracy of Apple services is significantly improved. Our work can provide an effective solution for the fine management of mobile encryption services.

Most existing deep neural networks (DNNs) are inexplicable and fragile, which can be easily deceived by carefully designed adversarial example with tiny undetectable noise. This allows attackers to cause serious consequences in many DNN-assisted scenarios without human perception. In the field of speaker recognition, the attack for speaker recognition system has been relatively mature. Most works focus on white-box attacks that assume the information of the DNN is obtainable, and only a few works study gray-box attacks. In this paper, we study blackbox attacks on the speaker recognition system, which can be applied in the real world since we do not need to know the system information. By combining the idea of transferable attack and query attack, our proposed method NMI-FGSM-Tri can achieve the targeted goal by misleading the system to recognize any audio as a registered person. Specifically, our method combines the Nesterov accelerated gradient (NAG), the ensemble attack and the restart trigger to design an attack method that generates the adversarial audios with good performance to attack blackbox DNNs. The experimental results show that the effect of the proposed method is superior to the extant methods, and the attack success rate can reach as high as 94.8% even if only one query is allowed.

Automatic speech recognition (ASR) models are used widely in applications for voice navigation and voice control of domestic appliances. ASRs have been misused by attackers to generate malicious outputs by attacking the deep learning component within ASRs. To assess the security and robustnesss of ASRs, we propose techniques within our framework SPAT that generate blackbox (agnostic to the DNN) adversarial attacks that are portable across ASRs. This is in contrast to existing work that focuses on whitebox attacks that are time consuming and lack portability. Our techniques generate adversarial attacks that have no human audible difference by manipulating the input speech signal using a psychoacoustic model that maintains the audio perturbations below the thresholds of human perception. We propose a framework SPAT with three attack generation techniques based on the psychoacoustic concept and frame selection techniques to selectively target the attack. We evaluate portability and effectiveness of our techniques using three popular ASRs and two input audio datasets using the metrics- Word Error Rate (WER) of output transcription, Similarity to original audio, attack Success Rate on different ASRs and Detection score by a defense system. We found our adversarial attacks were portable across ASRs, not easily detected by a state-of the-art defense system, and had significant difference in output transcriptions while sounding similar to original audio.

In recent years, radar automatic target recognition (RATR) technology based on high-resolution range profile (HRRP) has received extensive attention in various fields. However, insufficient data on non-cooperative targets seriously affects recognition performance of this technique. For HRRP target recognition under few-shot condition, we proposed a novel gaussian deep belief network based on model-agnostic meta-learning (GDBN-MAML). In the proposed method, GDBN allowed real-value data to be transmitted over the entire network, which effectively avoided feature loss due to binarization requirements of conventional deep belief network (DBN) for data. In addition, we optimized the initial parameters of GDBN by multi-task learning based on MAML. In this way, the number of training samples required by the model for new recognition tasks could be reduced. We applied the proposed method to the HRRP recognition experiments of 3 types of 3D simulated aircraft models. The experimental results showed that the proposed method had higher recognition accuracy and generalization performance under few-shot condition compared with conventional deep learning methods.

Redaction of personal, confidential and sensitive information from documents is becoming increasingly important for individuals and organizations. In past years, there have been many well-publicized cases of data leaks from various popular companies. When the data contains sensitive information, these leaks pose a serious threat. To protect and conceal sensitive information, many companies have policies and laws about processing and sanitizing sensitive information in business documents.The traditional approach of manually finding and matching millions of words and then redacting is slow and error-prone. This paper examines different models to automate the identification and redaction of personal and sensitive information contained within the documents using named entity recognition. Sensitive entities example person’s name, bank account details or Aadhaar numbers targeted for redaction, are recognized based on the file’s content, providing users with an interactive approach to redact the documents by changing selected sensitive terms.

Robustness, both to accident and to malevolent perturbations, is a crucial determinant of the successful deployment of deep convolutional neural network based SAR ATR systems in various security-sensitive applications. This paper performs a detailed adversarial robustness evaluation of deep convolutional neural network based SAR ATR models across two public available SAR target recognition datasets. For each model, seven different adversarial perturbations, ranging from gradient based optimization to self-supervised feature distortion, are generated for each testing image. Besides adversarial average recognition accuracy, feature attribution techniques have also been adopted to analyze the feature diffusion effect of adversarial attacks, which promotes the understanding of vulnerability of deep learning models.

With the increasingly extensive use of diver and unmanned underwater vehicle in military, it has posed a serious threat to the security of the national coastal area. In order to prevent the underwater diver's impact on the safety of water area, it is of great significance to identify underwater small targets in time to make early warning for it. In this paper, convolutional neural network is applied to underwater small target recognition. The recognition targets are diver, whale and dolphin. Due to the time-frequency spectrum can reflect the essential features of underwater target, convolutional neural network can learn a variety of features of the acoustic signal through the image processed by the time-frequency spectrum, time-frequency image is input to convolutional neural network to recognize the underwater small targets. According to the study of learning rate and pooling mode, the network parameters and structure suitable for underwater small target recognition in this paper are selected. The results of data processing show that the method can identify underwater small targets accurately.

The proliferation of mobile augmented reality applications and the toolkits to create them have serious implications for user privacy. In this paper, we explore how malicious AR app developers can leverage capabilities offered by commercially available AR libraries, and describe how edge computing can be used to address this privacy problem.

An enormous number of images are currently shared through social networking services such as Facebook. These images usually contain appearance of people and may violate the people's privacy if they are published without permission from each person. To remedy this privacy concern, visual privacy protection, such as blurring, is applied to facial regions of people without permission. However, in addition to image quality degradation, this may spoil the context of the image: If some people are filtered while the others are not, missing facial expression makes comprehension of the image difficult. This paper proposes an image melding-based method that modifies facial regions in a visually unintrusive way with preserving facial expression. Our experimental results demonstrated that the proposed method can retain facial expression while protecting privacy.