Author name cluster

Xiaohui Zhang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

9 papers

2 author rows

JBHI Journal 2026 Journal Article

Efficient Collaborative Model Training Mechanism With Privacy-Preserving Data for the IoMT

Chi Zhang
Tao Shen
Fenhua Bai
Xiaohui Zhang
Ziyuan Zhao

As time-series data from the Internet of Medical Things (IoMT) increasingly permeates various aspects of medical research, public governance, and clinical treatment, its sensitivity raises significant privacy concerns, hindering the potential of deep learning applications for cross-institutional data integration. Previous practices focused on deep learning methods based on centralized data storage and processing, which are often unsuitable for decentralized and privacy-sensitive IoMT data scenarios. Most existing methods rely on mechanisms such as trusted coordinators, which face challenges in addressing potential passive data leakage and side-channel attacks, failing to effectively protect the privacy of sensitive data during collaborative training. To address these issues, we propose a privacy-preserving collaborative training model, Secure Long Sequence Time-Series Forecasting (SecLSTF), for IoMT time-series data and design a mapping strategy between model components and Multi-Party Computation (MPC) protocols. Building on this foundation, we propose a novel secret sharing protocol, Pleione, which focuses on optimizing the computational efficiency of the low-level secret-sharing protocol. The protocol centers on a hyper-invertible matrix and adopts a paired double random expansion mechanism, significantly reducing the communication rounds required for random number generation. This optimization enhances the overall training speed of SecLSTF. Subsequently, we replace the original computational support protocol with Pleione. Experimental results show that SecLSTF-Pleione significantly reduces computational time while maintaining computational accuracy, outperforming other protocols in component efficiency. This study offers a potential pathway for cross-institutional IoMT data sharing.

Details DOI

AAAI Conference 2026 Conference Paper

FANoise: Singular Value-Adaptive Noise Modulation for Robust Multimodal Representation Learning

Jiaoyang Li
Jun Fang
Tianhao Gao
Xiaohui Zhang
Zhiyuan Liu
Chao Liu
Pengzhang Liu
Qixia Jiang

Representation learning is fundamental to modern machine learning, powering applications such as text retrieval and multimodal understanding. However, learning robust and generalizable representations remains challenging. While prior work has demonstrated that active noise injection, a form of data augmentation, can enhance encoding performance, most existing methods rely on heuristic or static noise, overlooking the dynamic nature of feature distributions during training. In this work, we systematically study the role of noise in representation learning from both gradient-based and feature distribution perspectives, using InfoNCE loss as a representative example. Focusing on multimodal representation learning, we propose FANoise, a novel feature-adaptive noise injection strategy. By leveraging the dynamics of contrastive learning, FANoise effectively mitigates the negative impacts of noise while preserving its benefits. Under this theoretically grounded framework, comprehensive experiments demonstrate that FANoise consistently improves overall performance on multimodal tasks across various base VLM models.

PDF Details DOI

YNIMG Journal 2025 Journal Article

A novel approach of [18F]FDG PET-based individual metabolic radiomics network to predict cognitive impairment in multiple system atrophy

Daoyan Hu
Xiaofeng Dou
Jing Wang
Chentao Jin
Ke Liu
Rui Zhou
Xiaohui Zhang
Congcong Yu

PURPOSE: F]FDG PET imaging to investigate brain metabolic connectivity patterns of MSA and validated the usefulness of IMRN-based predictive model for MSA-related cognitive impairment. METHODS: F]FDG PET/CT scans. IMRN was constructed by extracting non-redundant radiomics features from each brain region and computing pairwise Pearson correlation coefficients among these features. The validation of IMRN included assessments of small-world properties, test-retest reliability, and metabolic-genetic correlations. Connectome-based predictive modeling (CPM) was implemented to predict Mini Mental State Examination (MMSE) scores, while network-based statistics (NBS) were compared between MSA patients with cognitive impairment (MSA-CI, n = 58; MMSE 1), high reliability (average edge ICC = 0.754), and a significant correlation with gene expression (r = 0.44, P < 0.001). CPM significantly predicted cognitive scores through IMRN edges (positive network: r = 0.27, P = 0.03; negative network: r = 0.28, P = 0.02). NBS revealed decreased cerebellar-cortical connectivity (73 edges) and increased intra-cerebellar/limbic connectivity (24 edges) in MSA-CI compared to MSA-NC. The IMRN-based SVM outperformed SUVR-based SVM in classifying MSA-CI (accuracy: 73.91% vs 62.61%; AUC: 0.80 vs 0.69). CONCLUSION: This study established a novel approach of IMRN for assessing whole brain metabolic connectivity, uncovering distinct cerebellar connectivity patterns in MSA-CI, which held promise for facilitating personalized cognitive evaluations in MSA.

Details DOI

JBHI Journal 2025 Journal Article

Real-Time Continuous Locomotion Mode Recognition and Transition Prediction for Human With Lower Limb Exoskeleton

Xunju Ma
Yali Liu
Xiaohui Zhang
Lorenzo Masia
Qiuzhi Song

Real-time continuous locomotion mode recognition and seamless timely transition detection is critical for the exoskeleton robot. This study aims to present a comprehensive and innovative framework for locomotion mode recognition and transition prediction, exclusively utilizing inertial measurement unit (IMU) signals from the exoskeleton. In this framework, a CNN-BiLSTM model was developed and trained to be the classifier and a novel majority filter was designed to reduce the transition misjudgment rate. Moreover, a comprehensive evaluation system encompassing eight dimensions for the classifier, incorporating evaluation metrics specifically for transition misjudgment, was proposed. We collected locomotion motion data from six subjects wearing a rigid exoskeleton robot using six IMU sensors on the exoskeleton. The proposed method achieves a high level of recognition accuracy, with an overall average of 99. 58 $\%$ for the five steady locomotion modes (level ground walking (LG), stair ascent/descent (SA/SD), and ramp ascent/descent (RA/RD)) across six subjects following the transition decision. All transitions are recognizable, and the majority can be predicted in advance, with an average prediction time of 353 ms. Furthermore, the implementation of majority filter resulted in an average 87. 04 $\%$ reduction in the transition misjudgment rate among six subjects, thereby decreasing the average transition misjudgment rate to 4. 82 $\%$. Finally, the model was tested on a Jetson Nano to verify its real-time performance. The results presented above were obtained under the condition where either leg could function as the first transition leg and revealed that the developed system was capable of achieving precise locomotion mode recognition and timely transition prediction, with high real-time performance.

Details DOI

AAAI Conference 2025 Conference Paper

Region-Based Optimization in Continual Learning for Audio Deepfake Detection

Yujie Chen
Jiangyan Yi
Cunhang Fan
Jianhua Tao
Yong Ren
Siding Zeng
Chu Yuan Zhang
Xinrui Yan

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 model’s ability to learn more generalized discriminative features. Experimental results show our method achieves a 21.3 percent 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.

PDF Details DOI

ICLR Conference 2025 Conference Paper

Vevo: Controllable Zero-Shot Voice Imitation with Self-Supervised Disentanglement

Xueyao Zhang
Xiaohui Zhang
Kainan Peng
Zhenyu Tang
Vimal Manohar
Yingru Liu
Jeff Hwang
Dangna Li

The imitation of voice, targeted on specific speech attributes such as timbre and speaking style, is crucial in speech generation. However, existing methods rely heavily on annotated data, and struggle with effectively disentangling timbre and style, leading to challenges in achieving controllable generation, especially in zero-shot scenarios. To address these issues, we propose Vevo, a versatile zero-shot voice imitation framework with controllable timbre and style. Vevo operates in two core stages: (1) Content-Style Modeling: Given either text or speech's content tokens as input, we utilize an autoregressive transformer to generate the content-style tokens, which is prompted by a style reference; (2) Acoustic Modeling: Given the content-style tokens as input, we employ a flow-matching transformer to produce acoustic representations, which is prompted by a timbre reference. To obtain the content and content-style tokens of speech, we design a fully self-supervised approach that progressively decouples the timbre, style, and linguistic content of speech. Specifically, we adopt VQ-VAE as the tokenizer for the continuous hidden features of HuBERT. We treat the vocabulary size of the VQ-VAE codebook as the information bottleneck, and adjust it carefully to obtain the disentangled speech representations. Solely self-supervised trained on 60K hours of audiobook speech data, without any fine-tuning on style-specific corpora, Vevo matches or surpasses existing methods in accent and emotion conversion tasks. Additionally, Vevo’s effectiveness in zero-shot voice conversion and text-to-speech tasks further demonstrates its strong generalization and versatility. Audio samples are available at https://versavoice.github.io/.

Details

JMLR Journal 2024 Journal Article

Scaling Speech Technology to 1,000+ Languages

Vineel Pratap
Andros Tjandra
Bowen Shi
Paden Tomasello
Arun Babu
Sayani Kundu
Ali Elkahky
Zhaoheng Ni

Expanding the language coverage of speech technology has the potential to improve access to information for many more people. However, current speech technology is restricted to about one hundred languages which is a small fraction of the over 7,000 languages spoken around the world. The Massively Multilingual Speech (MMS) project increases the number of supported languages by 10-40x, depending on the task while providing improved accuracy compared to prior work. The main ingredients are a new dataset based on readings of publicly available religious texts and effectively leveraging self-supervised learning. We built pre-trained wav2vec 2.0 models covering 1,406 languages, a single multilingual automatic speech recognition model for 1,107 languages, speech synthesis models for the same number of languages, as well as a language identification model for 4,017 languages. Experiments show that our multilingual speech recognition model more than halves the word error rate of Whisper on 54 languages of the FLEURS benchmark while being trained on a small fraction of the labeled data. [abs] [ pdf ][ bib ] [ code ] &copy JMLR 2024. ( edit, beta )

PDF Details

AAAI Conference 2024 Conference Paper

What to Remember: Self-Adaptive Continual Learning for Audio Deepfake Detection

Xiaohui Zhang
Jiangyan Yi
Chenglong Wang
Chu Yuan Zhang
Siding Zeng
Jianhua Tao

The rapid evolution of speech synthesis and voice conversion has raised substantial concerns due to the potential misuse of such technology, prompting a pressing need for effective audio deepfake detection mechanisms. Existing detection models have shown remarkable success in discriminating known deepfake audio, but struggle when encountering new attack types. To address this challenge, one of the emergent effective approaches is continual learning. In this paper, we propose a continual learning approach called Radian Weight Modification (RWM) for audio deepfake detection. The fundamental concept underlying RWM involves categorizing all classes into two groups: those with compact feature distributions across tasks, such as genuine audio, and those with more spread-out distributions, like various types of fake audio. These distinctions are quantified by means of the in-class cosine distance, which subsequently serves as the basis for RWM to introduce a trainable gradient modification direction for distinct data types. Experimental evaluations against mainstream continual learning methods reveal the superiority of RWM in terms of knowledge acquisition and mitigating forgetting in audio deepfake detection. Furthermore, RWM's applicability extends beyond audio deepfake detection, demonstrating its potential significance in diverse machine learning domains such as image recognition.

PDF Details DOI

EAAI Journal 2023 Journal Article

KGCL: A Knowledge-enhanced Graph Contrastive learning framework for session-based recommendation

Xiaohui Zhang
Huifang Ma
Fanyi Yang
Zhixin Li
Liang Chang

Session-based Recommendation (SBR) aims at predicting the next item based on a short-term anonymous user behavior, whose main challenge lies at the sparsity problem of user–item interactions. Graph contrastive learning, which discovers ground-truth samples by data augmentation, is a promising paradigm to tackle this problem. However, the following two insights are often overlooked by most of these contrastive learning-based models. First, item knowledge (i. e. , item attributes, which can be distilled from open knowledge graphs) provides side information to model the complex high-order relations among items. Second, effective embedding aggregating mechanism is capable of filtering noisy preference signals (i. e. , unrelated items) in sessions and retaining higher weight for the related items. These insights motivate us to construct an item attribute hypergraph to summarize associations among items that share common attributes and develop a Knowledge-enhanced Graph Contrastive Learning framework for session-based recommendation (KGCL). Technically, KGCL constructs two independent and complementary views (cross-session graph and item attribute hypergraph) in terms of user–item interactions sequence and item intrinsic attributes respectively, so as to explicitly capture both internal and external factors of items. Then, we encode item and session embeddings with a query-aware graph attention network and a hypergraph convolutional network over the above two views. Finally, we devise two contrastive learning loss — global–global contrastive learning and local–global contrastive learning — that maximize agreement between these two views and generate high-quality recommendation results. Extensive experiments conducted on three real-world datasets show KGCL has a higher expressive power that enables SBR to predict the next item.

Details DOI