A5021697903- Emotion and Mood Recognition
- Speech and Audio Processing
- Spectroscopy Techniques in Biomedical and Chemical Research
- Sentiment Analysis and Opinion Mining
- Speech Recognition and Synthesis
- EEG and Brain-Computer Interfaces
- Multimodal Machine Learning Applications
- Music and Audio Processing
- Advanced Sensor and Control Systems
- Topic Modeling
- Neural and Behavioral Psychology Studies
- Wireless Sensor Networks and IoT
- Advanced Image and Video Retrieval Techniques
- Network Security and Intrusion Detection
- Anomaly Detection Techniques and Applications
- Face and Expression Recognition
- Face recognition and analysis
- Human Pose and Action Recognition
- Adversarial Robustness in Machine Learning
- Image and Signal Denoising Methods
- Bee Products Chemical Analysis
- Natural Language Processing Techniques
- Sparse and Compressive Sensing Techniques
- Speech and dialogue systems
- Technology and Security Systems
Zhejiang Lab
2020-2025
Ostfalia University of Applied Sciences
2022-2024
Institute of Art
2024
TU Wien
2023
Institut für Mikroelektronik- und Mechatronik-Systeme
2023
Otto-von-Guericke University Magdeburg
2023
Private Hochschule für Wirtschaft und Technik
2023
Technical University of Applied Sciences Wildau
2023
University Hospital Magdeburg
2023
University of Applied Sciences Ravensburg-Weingarten
2023
In this paper, we study an untouched problem in visible-infrared person re-identification (VI-ReID), namely, Twin Noise Labels (TNL) which refers to as noisy annotation and correspondence. brief, on the one hand, it is inevitable annotate some persons with wrong identity due complexity data collection annotation, e.g., poor recognizability infrared modality. On other wrongly annotated a single modality will eventually contaminate cross-modal correspondence, thus leading To solve TNL problem,...
Surface-enhanced Raman spectroscopy of serum accurately detects prostate cancer in patients with prostate-specific antigen levels 4–10 ng/mL Na Chen,1 Ming Rong,1 Xiaoguang Shao,2 Heng Zhang,1 Shupeng Liu,1,3 Baijun Dong,2 Wei Xue,2 Tingyun Wang,1 Taihao Li,3 Jiahua Pan2 1Key Laboratory Specialty Fiber Optics and Optical Access Networks, School Communication Information Engineering, Shanghai University, 2Department Urology, Ren Ji Hospital, Medicine, Jiao Tong Shanghai, 3Beijing...
3D dense captioning requires a model to translate its understanding of an input scene into several captions associated with different object regions. Existing methods adopt sophisticated "detect-then-describe" pipeline, which builds explicit relation modules upon detector numerous hand-crafted components. While these have achieved initial success, the cascade pipeline tends accumulate errors because duplicated and inaccurate box estimations messy scenes. In this paper, we first propose...
Abstract Inspired by biological processes, feature learning techniques, such as deep learning, have achieved great success in various fields. However, since organs may operate differently from semiconductor devices, models usually require dedicated hardware and are computation-complex. High energy consumption has made model growth unsustainable. We present an approach that directly implements using physics to minimize disparity between hardware. Following this approach, a technique based on...
Affective computing is a rapidly growing multidisciplinary field that encompasses computer science, engineering, psychology, neuroscience, and other related disciplines. Although the literature in this has progressively grown matured, lack of comprehensive bibliometric analysis limits overall understanding theory, technical methods, applications affective computing. This review presents quantitative 33,448 articles published period from 1997 to 2023, identifying challenges, calling attention...
In this work, we presented a deep learning approach based on the LeNet-5 network for analysing and classifying recognisable shapes in urine sample images. The is shape analysis to recognise classify red blood cells, white epithelial cells crystals observed under microscopes samples. We modified neural by changing numbers of output nodes convolutional layers. compared results our method with those obtained traditional feature extraction followed back-propagation networks. Our testing showed...
Early diagnosis of prostate cancer (PCa) is always a great challenge in clinical practice, especially distinguishing benign prostatic hyperplasia (BPH) from early cancer, due to the high similarity pathology prostate‐specific antigen (PSA) test and radiological detection. The conventional diagnostic methods are often less efficient specificity accuracy, leading quite few unnecessary biopsies. This work establishes noninvasive method for PCa by investigating urine samples using Raman...
Jun Sun, Shoukang Han, Yu-Ping Ruan, Xiaoning Zhang, Shu-Kai Zheng, Yulong Liu, Yuxin Huang, Taihao Li. Proceedings of the 61st Annual Meeting Association for Computational Linguistics (Volume 1: Long Papers). 2023.
Multimodal learning is susceptible to modality missing, which poses a major obstacle for its practical applications and, thus, invigorates increasing research interest. In this paper, we investigate two challenging problems: 1) when missing exists in the training data, how exploit incomplete samples while guaranteeing that they are properly supervised? 2) rates of different modalities vary, causing or exacerbating imbalance among modalities, address and ensure all well-trained. To tackle...
In this work we presented a new parameter-free thresholding method for image segmentation. separating an into two classes, the employs objective function that not only maximizes between-class variance but also distance between mean of each class and global image. The design aims to circumvent challenge many existing techniques encounter when underlying classes have very different sizes or variances. Advantages are two-fold. First, it is parameter-free, meaning can generate consistent...
Prosody plays a fundamental role in human speech and communication, facilitating intelligibility conveying emotional cognitive states. Extracting accurate prosodic information from is vital for building assistive technology, such as controllable synthesis, speaking style transfer, emotion recognition (SER). However, it challenging to disentangle speaker-independent prosody representations since attributes, intonation, excessively entangle with speaker-specific e.g., pitch. In this article,...
Using recombinase-mediated cassette exchange to test multiple transgenes at the same site of integration, we demonstrate a novel chromatin context-dependent silencer activity β-globin locus control region (LCR). This requires DNase I hypersensitive sites HS2 and HS3 but not HS4. After silencing, silenced cassettes adopt typical closed conformation (histone H3 H4 deacetylation, histone H3-K4 methylation, DNA replication in late S phase). In absence LCR remains decondensed. We that is...
Multi-modal emotion recognition (MER) using speech and text has attracted extensive attention because of the easy availability data for these two modalities. Recently, self-surprised learning (SSL) pre-trained model become state-of-the-art (SOTA) method extraction acoustic textual features. However, SSL representation may lose some important paralinguistic information, resulting in limited knowledge MER. In this paper, we propose to adopt kinds features (i.e., spectral feature) as inputs...