Advances in image generation enable hyper-realistic synthetic faces but also pose risks, thus making synthetic face detection crucial. Previous research focuses on the general differences between generated images and real images, often overlooking the discrepancies among various generative techniques. In this paper, we explore the intrinsic relationship between synthetic images and their corresponding generation technologies. We find that specific images exhibit significant reconstruction discrepancies across different generative methods and that matching generation techniques provide more accurate reconstructions. Based on this insight, we propose a Multi-Reconstruction-based detector. By reversing and reconstructing images using multiple generative models, we analyze the reconstruction differences among real, GAN-generated, and DM-generated images to facilitate effective differentiation. Additionally, we introduce the Asian Synthetic Face Dataset (ASFD), containing synthetic Asian faces generated with various GANs and DMs. This dataset complements existing synthetic face datasets. Experimental results demonstrate that our detector achieves exceptional performance, with strong generalization and robustness.

Detecting early ischemic lesions (EIL) in computed tomography (CT) images is crucial for reducing diagnostic time and minimizing neuron loss due to oxygen deprivation. This paper introduces DCTP-Net, a dual-branch network for segmenting acute ischemic stroke lesions in CT images, consisting of a segmentation branch and a prompt-aware branch. The segmentation branch uses an encoder-decoder network as the backbone to identify lesions, where the encoder fuses CT image features with prompt features from the prompt-aware branch. To enhance semantic feature extraction and reduce the impact of cerebral structural details, we introduce a cross-collaboration dynamic connection (CCDC) module to link the encoder and decoder. The prompt-aware branch includes a learnable prompt (LP) block to incorporate cerebral prior knowledge, and the prompt-aware encoder (PAE) combines the LP block with multi-level features from the segmentation branch for more precise representation. Additionally, we propose a CLIP-enhance textual prompt (CETP) module that utilizes the CLIP text encoder to generate specialized convolutional parameters for the segmentation head. These parameters are tailored to the unique characteristics of each input image, improving segmentation performance. Qualitative and quantitative studies reveal that DCTP-Net outperforms the current state-of-the-art, IS-Net, with Dice score increases of 3.9% on AISD and 3.8% on ISLES2018, demonstrating its superiority in EIL segmentation.

Recently, domain adaptation model was studied for Speaker Verification (SV) task to solve the performance reduction resulted from various mismatch between the training dataset and the testing dataset. However, since most schemes apply adaptation from single source domain to the target domain, they may not be robust enough against various mismatch between the samples in the source domain and those in the target domain. In this paper, multiple source domain adaptation and fusion model was studied to solve this problem. First, the x-vector based SV scheme is pretrained on each source domain. Second, UnShared Network (USN) combining weight regularization loss based adaptation model is adopted to adapt each pretrained model to the target domain and maintain the speaker-related feature as much as possible. Third, the Registered-Utterance-To-Test-Utterance (RUTTU) similarity matrix is constructed based on the x-vector feature extracted by the adapted model from each source domain. Fourth, the Similarity Network Fusion (SNF) technique is introduced and modified to obtain Modified Similarity Network Fusion (MSNF) scheme, which is used to fuse the RUTTU similarity matrices obtained from multiple adapted models. Finally, speaker matching is achieved by averaging the obtained fused similarities between the test utterance and each of the registered utterances of a specific speaker. Extensive experimental results demonstrate that i) the proposed fusion model outperforms the Multi-source Distilling Domain Adaptation (MDDA) model and Logistic Regression Fusion (LRF) model in SV task; ii) the USN combining weight regularization loss based adaptation strategy, the MSNF fusion scheme, and the modified speaker matching mechanism all contribute to the performance enhancement; iii) the fusion effectiveness is little influenced by the hyper-parameters setting of MSNF.

By providing customers with conveniences such as easy access to an extensive variety of restaurants, effortless food ordering and fast delivery, on-demand food delivery (OFD) platforms have achieved explosive growth in recent years. A crucial machine learning task performed at OFD platforms is prediction of the Order Fulfillment Cycle Time (OFCT), which refers to the amount of time elapsed between a customer places an order and he/she receives the meal. The accuracy of predicted OFCT is important for customer satisfaction, as it needs to be communicated to a customer before he/she places the order, and is considered as a service promise that should be fulfilled as well as possible. As a result, the estimated OFCT also heavily influences planning decisions such as dispatching and routing.

Generalized likelihood ratio test is an effective method for target detection in passive radar systems. The distribution of its decision variable in the presence of a direct path is unknown but is required for the calculation of the detection threshold and the detection probability. In this letter, several new approximations to this distribution are proposed by using moment matching. Numerical results show that the generalized extreme value approximation works consistently well for both null and alternative hypotheses with large or small signal-to-noise ratios. On the other hand, the Gaussian and logistic approximations only work well in the null hypothesis.

Self-similarity analysis-based feature summarizing technique (SuCo) was proposed recently to improve the time and memory efficiency of Cover Song Identification (CSI). In this paper, both the feature summarizing and the cross-similarity calculating strategies of the SuCo model are modified as follows to enhance its identification accuracy. At the feature summarizing stage, first, the Hubness Reduction (HR) strategy is adopted to reduce the possible `Hubness' phenomenon existing in the feature subsequence community, which may affect the retrieval effectiveness. Then, the Network Enhancement (NE) technique, which was originally proposed in biology to improve gene-function prediction accuracy, is introduced to reduce the noise in the self-similarity network caused by the limitation of feature extraction and similarity measuring, and the inherent musical and acoustic variations. At the cross-similarity calculating stage, first, the summarized representative feature subsequences of the reference are concatenated to obtain its combined representative feature. Then, considering that the nonlinear recurrence property is important for describing the melody perception-based similarity, Qmax is adopted to measure the similarity between the combined representative feature of the reference and the unsummarized feature sequence of the query. Extensive experiments carried out on four open CSI datasets with 5 types of features and 2 kinds of representative feature subsequence choosing methods verify that: i) The proposed scheme outperforms the SuCo model in retrieval effectiveness. ii) Each of the above modifications contributes to the performance enhancement of the proposed scheme. iii) The proposed scheme achieves high generalization.

Fuzzy reasoning analysis (FRA) is known as a mathematical programming approach to strong nonlinear system prediction. In this paper, the design and implementation of a fuzzy reasoning based cyberspace security awareness model is presented. The new feature of the this method is that it can give better perception and projection of cyberspace security situational. We show that this new method, referred to as machine learning situational awareness analyses, is efficient in situational perceiving, main contribution of this paper is to further integration of the fuzzy reasoning and situational risk control strategies to cyberspace security analysis.

Emergencies and disruption events bring new challenges to disruption management, a new analysis method is presented in this paper to search the solution, through the analysis of situational awareness, reasons of trend and changes of disruption events in different stages get the impact to whole supply chain become clear, this paper also provides the idea of situational awareness indexes system construction, and some rules should be obey when design the index system. At last, “bullwhip effect” is adopted as an example to analyze how the situational awareness information is delayed.

To make digital audio watermarking accomplish both copyright protection and content authentication with localization, a novel multipurpose watermarking scheme is proposed in this paper. The fragile watermark (for content authentication) is embedded in the detail wavelet coefficients based on instantaneous mixing model of independent component analysis. While, the robust watermark (for copyright protection) is embedded in the secret key based on the essential features of the host audio, which are extracted based on the combination of discrete wavelet transform, discrete cosine transform and higher-order cumulant. Simulation results demonstrate the effectiveness of our algorithm in terms of transparency, robustness, detection reliability and tampering localization.