Jun Xue

Jun Xue, Cunhang Fan, Zhao Lv et al.

Recently, pioneer research works have proposed a large number of acoustic features (log power spectrogram, linear frequency cepstral coefficients, constant Q cepstral coefficients, etc.) for audio deepfake detection, obtaining good performance, and showing that different subbands have different contributions to audio deepfake detection. However, this lacks an explanation of the specific information in the subband, and these features also lose information such as phase. Inspired by the mechanism of synthetic speech, the fundamental frequency (F0) information is used to improve the quality of synthetic speech, while the F0 of synthetic speech is still too average, which differs significantly from that of real speech. It is expected that F0 can be used as important information to discriminate between bonafide and fake speech, while this information cannot be used directly due to the irregular distribution of F0. Insteadly, the frequency band containing most of F0 is selected as the input feature. Meanwhile, to make full use of the phase and full-band information, we also propose to use real and imaginary spectrogram features as complementary input features and model the disjoint subbands separately. Finally, the results of F0, real and imaginary spectrogram features are fused. Experimental results on the ASVspoof 2019 LA dataset show that our proposed system is very effective for the audio deepfake detection task, achieving an equivalent error rate (EER) of 0.43%, which surpasses almost all systems.

Jun Xue, Cunhang Fan, Jiangyan Yi et al.

In this paper, we propose a novel self-distillation method for fake speech detection (FSD), which can significantly improve the performance of FSD without increasing the model complexity. For FSD, some fine-grained information is very important, such as spectrogram defects, mute segments, and so on, which are often perceived by shallow networks. However, shallow networks have much noise, which can not capture this very well. To address this problem, we propose using the deepest network instruct shallow network for enhancing shallow networks. Specifically, the networks of FSD are divided into several segments, the deepest network being used as the teacher model, and all shallow networks become multiple student models by adding classifiers. Meanwhile, the distillation path between the deepest network feature and shallow network features is used to reduce the feature difference. A series of experimental results on the ASVspoof 2019 LA and PA datasets show the effectiveness of the proposed method, with significant improvements compared to the baseline.

Jiayu Xiong, Jing Wang, Qi Zhang et al.

Real-world multimodal systems must be robust against low-quality data, such as sensor noise, incomplete multimodal data and conflicting inputs. However, existing trustworthy fusion methods rely on the model's own prediction confidence to judge data quality. This creates a circular dependency: when a model is confident but wrong, these methods fail to detect the error. To break this loop, we propose Geometry-based Multimodal Fusion (GMF). Instead of relying on predictions, we evaluate reliability by measuring how much transport correction the input needs in latent space. We implement Diffusion Schrödinger Bridge transport with Rectified Flow, where the squared initial velocity gives an efficient learned correction score. Valid data has low squared velocity magnitude, while noisy, incomplete data or conflicting data requires stronger transport correction. This geometry-based reliability signal acts as an independent judge, effectively flagging unreliable inputs even when the classifier is fooled. Extensive experiments demonstrate that GMF significantly improves robustness against severe sensor noise and semantic conflicts compared to confidence-based baselines.

Cunhang Fan, Hongyu Zhang, Wei Huang et al.

Auditory Attention Detection (AAD) aims to detect target speaker from brain signals in a multi-speaker environment. Although EEG-based AAD methods have shown promising results in recent years, current approaches primarily rely on traditional convolutional neural network designed for processing Euclidean data like images. This makes it challenging to handle EEG signals, which possess non-Euclidean characteristics. In order to address this problem, this paper proposes a dynamical graph self-distillation (DGSD) approach for AAD, which does not require speech stimuli as input. Specifically, to effectively represent the non-Euclidean properties of EEG signals, dynamical graph convolutional networks are applied to represent the graph structure of EEG signals, which can also extract crucial features related to auditory spatial attention in EEG signals. In addition, to further improve AAD detection performance, self-distillation, consisting of feature distillation and hierarchical distillation strategies at each layer, is integrated. These strategies leverage features and classification results from the deepest network layers to guide the learning of shallow layers. Our experiments are conducted on two publicly available datasets, KUL and DTU. Under a 1-second time window, we achieve results of 90.0\% and 79.6\% accuracy on KUL and DTU, respectively. We compare our DGSD method with competitive baselines, and the experimental results indicate that the detection performance of our proposed DGSD method is not only superior to the best reproducible baseline but also significantly reduces the number of trainable parameters by approximately 100 times.

Jun Xue, Tong Zhang, Zhuolin Yi et al.

The rapid advancement of generative AI has made audio deepfakes increasingly indistinguishable from authentic human vocals, posing significant threats to persons-of-interest (POI) such as public figures. Current detection systems primarily rely on generic, black-box models that fail to capture speaker-specific idiosyncratic traits and lack interpretability. In this paper, we propose Phoneme-based Voice Profiling (PVP), a novel personalized defense framework. By shifting the detection paradigm from macro-utterance analysis to micro-phonetic modeling, PVP captures the unique acoustic distributions underlying a POI's habitual articulatory patterns. Specifically, our framework models speaker-specific phonetic realizations using lightweight Gaussian Mixture Models (GMMs) estimated solely from bona fide reference speech. This design enables data-efficient profiling and robust generalization to previously unseen spoofing attacks without requiring heavy spoof-specific training. Furthermore, we introduce the first large-scale Chinese POI deepfake dataset to benchmark speaker-specific detection. Experimental results demonstrate that PVP significantly outperforms state-of-the-art generic detectors in POI spoofing scenarios, achieving substantial EER reductions while providing fine-grained, phoneme-level interpretability for forensic analysis. Code and data are available at: https://github.com/JunXue-tech/PVP

Shunbo Dong, Jun Xue, Cunhang Fan et al.

In this paper, we propose the multi-perspective information fusion (MPIF) Res2Net with random Specmix for fake speech detection (FSD). The main purpose of this system is to improve the model's ability to learn precise forgery information for FSD task in low-quality scenarios. The task of random Specmix, a data augmentation, is to improve the generalization ability of the model and enhance the model's ability to locate discriminative information. Specmix cuts and pastes the frequency dimension information of the spectrogram in the same batch of samples without introducing other data, which helps the model to locate the really useful information. At the same time, we randomly select samples for augmentation to reduce the impact of data augmentation directly changing all the data. Once the purpose of helping the model to locate information is achieved, it is also important to reduce unnecessary information. The role of MPIF-Res2Net is to reduce redundant interference information. Deceptive information from a single perspective is always similar, so the model learning this similar information will produce redundant spoofing clues and interfere with truly discriminative information. The proposed MPIF-Res2Net fuses information from different perspectives, making the information learned by the model more diverse, thereby reducing the redundancy caused by similar information and avoiding interference with the learning of discriminative information. The results on the ASVspoof 2021 LA dataset demonstrate the effectiveness of our proposed method, achieving EER and min-tDCF of 3.29% and 0.2557, respectively.

Yubiao Yue, Jun Xue, Haihuang Liang et al.

Due to the lack of effective mpox detection tools, the mpox virus continues to spread worldwide and has once again been declared a public health emergency of international concern by the World Health Organization. Lightweight deep learning model-based detection systems are crucial to alleviate mpox outbreaks since they are suitable for widespread deployment, especially in resource-limited scenarios. However, the key to its successful application depends on ensuring that the model can effectively model local features and long-range dependencies in mpox lesions while maintaining lightweight. Inspired by the success of Mamba in modeling long-range dependencies and its linear complexity, we proposed a lightweight hybrid architecture called MpoxMamba for efficient mpox detection. MpoxMamba utilizes depth-wise separable convolutions to extract local feature representations in mpox skin lesions and greatly enhances the model's ability to model the global contextual information by grouped Mamba modules. Notably, MpoxMamba's parameter size and FLOPs are 0.77M and 0.53G, respectively. Experimental results on two widely recognized benchmark datasets demonstrate that MpoxMamba outperforms state-of-the-art lightweight models and existing mpox detection methods. Importantly, we developed a web-based online application to provide free mpox detection (http://5227i971s5.goho.co:30290). The source codes of MpoxMamba are available at https://github.com/YubiaoYue/MpoxMamba.

Jiayu Xiong, Jing Wang, Qi Zhang et al.

Audio-visual deepfake localization demands interval-level outputs that serve as temporal evidence. Despite recent progress, symmetric fusion under single-sided or asynchronous forgeries propagates cross-modal noise, degrading high-precision localization. We present IaMSB, an inconsistency-aware multimodal Schrödinger Bridge (SB) that jointly estimates cross-modal consistency and performs interval-level localization. Unlike diffusion models, SB minimizes path-distribution discrepancy and yields consistency scores without explicit noise injection or denoising. With the Schrödinger Bridge (SB), IaMSB unifies consistency estimation, cross-modal information selection, and bridge-step scheduling in one framework. Specifically, a lightweight coarse bridge first proposes candidate intervals and estimates cross-modal consistency; these statistics select cross-modal witness signals and allocate bridge steps asymmetrically across modalities. A refinement bridge then performs step-tuned fusion and outputs refined, time-aligned intervals. IaMSB anticipates single-sided and asynchronous forgeries and, using bottlenecked cross-modal interaction with step allocation, suppresses noise transfer, avoids unnecessary iterations. Across benchmarks, IaMSB stabilizes strict-IoU boundary precision, raising AP@0.95 by 3%~10%, and yields improved high-precision localization, particularly for single-sided forgeries.

Yihuan Huang, Jun Xue, Liu Jiajun et al.

Audio-Visual Speech Recognition (AVSR) has achieved remarkable progress in offline conditions, yet its robustness in real-world video conferencing (VC) remains largely unexplored. This paper presents the first systematic evaluation of state-of-the-art AVSR models across mainstream VC platforms, revealing severe performance degradation caused by transmission distortions and spontaneous human hyper-expression. To address this gap, we construct \textbf{MLD-VC}, the first multimodal dataset tailored for VC, comprising 31 speakers, 22.79 hours of audio-visual data, and explicit use of the Lombard effect to enhance human hyper-expression. Through comprehensive analysis, we find that speech enhancement algorithms are the primary source of distribution shift, which alters the first and second formants of audio. Interestingly, we find that the distribution shift induced by the Lombard effect closely resembles that introduced by speech enhancement, which explains why models trained on Lombard data exhibit greater robustness in VC. Fine-tuning AVSR models on MLD-VC mitigates this issue, achieving an average 17.5% reduction in CER across several VC platforms. Our findings and dataset provide a foundation for developing more robust and generalizable AVSR systems in real-world video conferencing. MLD-VC is available at https://huggingface.co/datasets/nccm2p2/MLD-VC.

Jun Xue, Zhuolin Yi, Yihuan Huang et al.

With the rapid advancement of speech generation technologies, the threat posed by speech deepfakes in real-time communication (RTC) scenarios has intensified. However, existing detection studies mainly focus on offline simulations and struggle to cope with the complex distortions introduced during RTC transmission, including unknown speech enhancement processes (e.g., noise suppression) and codec compression. To address this challenge, we present the first large-scale speech deepfake dataset tailored for RTC scenarios, termed \textit{RTCFake}, totaling approximately 600 hours. The dataset is constructed by transmitting speech through multiple mainstream social media and conferencing platforms (e.g., Zoom), enabling precise pairing between offline and online speech. In addition, we propose a phoneme-guided consistency learning (PCL) strategy that enforces models to learn platform-invariant semantic structural representations. In this paper, the RTCFake dataset is divided into training, development, and evaluation sets. The evaluation set further includes both unseen RTC platforms and unseen complex noise conditions, thereby providing a more realistic and challenging evaluation benchmark for speech deepfake detection. Furthermore, the proposed PCL strategy achieves significant improvements in both cross-platform generalization and noise robustness, offering an effective and generalizable modeling paradigm. The \textit{RTCFake} dataset is provided in the {https://huggingface.co/datasets/JunXueTech/RTCFake}.

Yujie Chen, Jiangyan Yi, Cunhang Fan et al.

Rapid advancements in speech synthesis and voice conversion bring convenience but also new security risks, creating an urgent need for effective audio deepfake detection. Although current models perform well, their effectiveness diminishes when confronted with the diverse and evolving nature of real-world deepfakes. To address this issue, we propose a continual learning method named Region-Based Optimization (RegO) for audio deepfake detection. Specifically, we use the Fisher information matrix to measure important neuron regions for real and fake audio detection, dividing them into four regions. First, we directly fine-tune the less important regions to quickly adapt to new tasks. Next, we apply gradient optimization in parallel for regions important only to real audio detection, and in orthogonal directions for regions important only to fake audio detection. For regions that are important to both, we use sample proportion-based adaptive gradient optimization. This region-adaptive optimization ensures an appropriate trade-off between memory stability and learning plasticity. Additionally, to address the increase of redundant neurons from old tasks, we further introduce the Ebbinghaus forgetting mechanism to release them, thereby promoting the capability of the model to learn more generalized discriminative features. Experimental results show our method achieves a 21.3% improvement in EER over the state-of-the-art continual learning approach RWM for audio deepfake detection. Moreover, the effectiveness of RegO extends beyond the audio deepfake detection domain, showing potential significance in other tasks, such as image recognition. The code is available at https://github.com/cyjie429/RegO

Yubiao Yue, Jun Xue, Haihua Liang et al.

Booming deep learning has substantially improved the diagnosis for diverse lesions in ultrasound images, but a conspicuous research gap concerning cervical lymph node lesions still remains. The objective of this work is to diagnose cervical lymph node lesions in ultrasound images by leveraging a deep learning model. To this end, we first collected 3392 cervical ultrasound images containing normal lymph nodes, benign lymph node lesions, malignant primary lymph node lesions, and malignant metastatic lymph node lesions. Given that ultrasound images are generated by the reflection and scattering of sound waves across varied bodily tissues, we proposed the Conv-FFT Block. It integrates convolutional operations with the fast Fourier transform to more astutely model the images. Building upon this foundation, we designed a novel architecture, named SFUSNet. SFUSNet not only discerns variances in ultrasound images from the spatial domain but also adeptly captures micro-structural alterations across various lesions in the frequency domain. To ascertain the potential of SFUSNet, we benchmarked it against 12 popular architectures through five-fold cross-validation. The results show that SFUSNet is the state-of-the-art model and can achieve 92.89% accuracy. Moreover, its average precision, average sensitivity and average specificity for four types of lesions achieve 90.46%, 89.95% and 97.49%, respectively.

Yubiao Yue, Jun Xue, Chao Wang et al.

Numerous studies have affirmed that deep learning models can facilitate early diagnosis of lesions in endoscopic images. However, the lack of available datasets stymies advancements in research on nasal endoscopy, and existing models fail to strike a good trade-off between model diagnosis performance, model complexity and parameters size, rendering them unsuitable for real-world application. To bridge these gaps, we created the first large-scale nasal endoscopy dataset, named 7-NasalEID, comprising 11,352 images that contain six common nasal diseases and normal samples. Subsequently, we proposed U-SEANNet, an innovative U-shaped architecture, underpinned by depth-wise separable convolution. Moreover, to enhance its capacity for detecting nuanced discrepancies in input images, U-SEANNet employs the Global-Local Channel Feature Fusion module, enabling it to utilize salient channel features from both global and local contexts. To demonstrate U-SEANNet's potential, we benchmarked U-SEANNet against seventeen modern architectures through five-fold cross-validation. The experimental results show that U-SEANNet achieves a commendable accuracy of 93.58%. Notably, U-SEANNet's parameters size and GFLOPs are only 0.78M and 0.21, respectively. Our findings suggest U-SEANNet is the state-of-the-art model for nasal diseases diagnosis in endoscopic images.

Shuo Gao, Jingyang Zhang, Jun Xue et al.

Carotid atherosclerosis represents a significant health risk, with its early diagnosis primarily dependent on ultrasound-based assessments of carotid intima-media thickening. However, during carotid ultrasound screening, significant view variations cause style shifts, impairing content cues related to thickening, such as lumen anatomy, which introduces spurious correlations that hinder assessment. Therefore, we propose a novel causal-inspired method for assessing carotid intima-media thickening in frame-wise ultrasound videos, which focuses on two aspects: eliminating spurious correlations caused by style and enhancing causal content correlations. Specifically, we introduce a novel Spurious Correlation Elimination (SCE) module to remove non-causal style effects by enforcing prediction invariance with style perturbations. Simultaneously, we propose a Causal Equivalence Consolidation (CEC) module to strengthen causal content correlation through adversarial optimization during content randomization. Simultaneously, we design a Causal Transition Augmentation (CTA) module to ensure smooth causal flow by integrating an auxiliary pathway with text prompts and connecting it through contrastive learning. The experimental results on our in-house carotid ultrasound video dataset achieved an accuracy of 86.93\%, demonstrating the superior performance of the proposed method. Code is available at \href{https://github.com/xielaobanyy/causal-imt}{https://github.com/xielaobanyy/causal-imt}.

Jun Xue, Yi Chai, Yanzhen Ren et al.

Speech editing achieves semantic inversion by performing fine-grained segment-level manipulation on original utterances, while preserving global perceptual naturalness. Existing detection studies mainly focus on manually edited speech with explicit splicing artifacts, and therefore struggle to cope with emerging end-to-end neural speech editing techniques that generate seamless acoustic transitions. To address this challenge, we first construct a large-scale bilingual dataset, AiEdit, which leverages large language models to drive precise semantic tampering logic and employs multiple advanced neural speech editing methods for data synthesis, thereby filling the gap of high-quality speech editing datasets. Building upon this foundation, we propose PELM (Prior-Enhanced Audio Large Language Model), the first large-model framework that unifies speech editing detection and content localization by formulating them as an audio question answering task. To mitigate the inherent forgery bias and semantic-priority bias observed in existing audio large models, PELM incorporates word-level probability priors to provide explicit acoustic cues, and further designs a centroid-aggregation-based acoustic consistency perception loss to explicitly enforce the modeling of subtle local distribution anomalies. Extensive experimental results demonstrate that PELM significantly outperforms state-of-the-art methods on both the HumanEdit and AiEdit datasets, achieving equal error rates (EER) of 0.57\% and 9.28\% (localization), respectively.

Daixian Li, Jun Xue, Yanzhen Ren et al.

Recent advances in speech synthesis and voice conversion have greatly improved the naturalness and authenticity of generated audio. Meanwhile, evolving encoding, compression, and transmission mechanisms on social media platforms further obscure deepfake artifacts. These factors complicate reliable detection in real-world environments, underscoring the need for representative evaluation benchmarks. To this end, we introduce ML-ITW (Multilingual In-The-Wild), a multilingual dataset covering 14 languages, seven major platforms, and 180 public figures, totaling 28.39 hours of audio. We evaluate three detection paradigms: end-to-end neural models, self-supervised feature-based (SSL) methods, and audio large language models (Audio LLMs). Experimental results reveal significant performance degradation across diverse languages and real-world acoustic conditions, highlighting the limited generalization ability of existing detectors in practical scenarios. The ML-ITW dataset is publicly available.

Jun Xue, Junze Wang, Xinming Zhang et al.

Scaling Maximum Entropy Reinforcement Learning (RL) to high-dimensional humanoid control remains a formidable challenge, as the ``curse of dimensionality'' induces severe exploration inefficiency and training instability in expansive action spaces. Consequently, recent high-throughput paradigms have largely converged on deterministic policy gradients combined with massive parallel simulation. We challenge this compromise with FastDSAC, a framework that effectively unlocks the potential of maximum entropy stochastic policies for complex continuous control. We introduce Dimension-wise Entropy Modulation (DEM) to dynamically redistribute the exploration budget and enforce diversity, alongside a continuous distributional critic tailored to ensure value fidelity and mitigate high-dimensional value overestimation. Extensive evaluations on HumanoidBench and other continuous control tasks demonstrate that rigorously designed stochastic policies can consistently match or outperform deterministic baselines, achieving notable gains of 180\% and 400\% on the challenging \textit{Basketball} and \textit{Balance Hard} tasks.

Yihuan Huang, Jiajun Liu, Yanzhen Ren et al.

Recent talking head synthesis works typically adopt speech features extracted from large-scale pre-trained acoustic models. However, the intrinsic many-to-many relationship between speech and lip motion causes phoneme-viseme alignment ambiguity, leading to inaccurate and unstable lips. To further improve lip sync accuracy, we propose PASE (Phoneme-Aware Speech Encoder), a novel speech representation model that bridges the gap between phonemes and visemes. PASE explicitly introduces phoneme embeddings as alignment anchors and employs a contrastive alignment module to enhance the discriminability between corresponding audio-visual pairs. In addition, a prediction and reconstruction task is designed to improve robustness under noise and partial modality absence. Experimental results show PASE significantly improves lip sync accuracy and achieves state-of-the-art performance across both NeRF- and 3DGS-based rendering frameworks, outperforming conventional methods based on acoustic features by 13.7 % and 14.2 %, respectively. Importantly, PASE can be seamlessly integrated into diverse talking head pipelines to improve the lip sync accuracy without architectural modifications.

Jiayu Xiong, Jing Wang, Hengjing Xiang et al.

Previous studies have highlighted significant advancements in multimodal fusion. Nevertheless, such methods often encounter challenges regarding the efficacy of feature extraction, data integrity, consistency of feature dimensions, and adaptability across various downstream tasks. This paper proposes a generalized multimodal fusion method (GMF) via the Poisson-Nernst-Planck (PNP) equation, which adeptly addresses the aforementioned issues. Theoretically, the optimization objective for traditional multimodal tasks is formulated and redefined by integrating information entropy and the flow of gradient backward step. Leveraging these theoretical insights, the PNP equation is applied to feature fusion, rethinking multimodal features through the framework of charged particles in physics and controlling their movement through dissociation, concentration, and reconstruction. Building on these theoretical foundations, GMF disassociated features which extracted by the unimodal feature extractor into modality-specific and modality-invariant subspaces, thereby reducing mutual information and subsequently lowering the entropy of downstream tasks. The identifiability of the feature's origin enables our approach to function independently as a frontend, seamlessly integrated with a simple concatenation backend, or serve as a prerequisite for other modules. Experimental results on multiple downstream tasks show that the proposed GMF achieves performance close to the state-of-the-art (SOTA) accuracy while utilizing fewer parameters and computational resources. Furthermore, by integrating GMF with advanced fusion methods, we surpass the SOTA results.

Cunhang Fan, Yujie Chen, Jun Xue et al.

In recent years, knowledge graph completion (KGC) models based on pre-trained language model (PLM) have shown promising results. However, the large number of parameters and high computational cost of PLM models pose challenges for their application in downstream tasks. This paper proposes a progressive distillation method based on masked generation features for KGC task, aiming to significantly reduce the complexity of pre-trained models. Specifically, we perform pre-distillation on PLM to obtain high-quality teacher models, and compress the PLM network to obtain multi-grade student models. However, traditional feature distillation suffers from the limitation of having a single representation of information in teacher models. To solve this problem, we propose masked generation of teacher-student features, which contain richer representation information. Furthermore, there is a significant gap in representation ability between teacher and student. Therefore, we design a progressive distillation method to distill student models at each grade level, enabling efficient knowledge transfer from teachers to students. The experimental results demonstrate that the model in the pre-distillation stage surpasses the existing state-of-the-art methods. Furthermore, in the progressive distillation stage, the model significantly reduces the model parameters while maintaining a certain level of performance. Specifically, the model parameters of the lower-grade student model are reduced by 56.7\% compared to the baseline.