A5079582557- Speech and Audio Processing
- Music and Audio Processing
- Speech Recognition and Synthesis
- Water resources management and optimization
- Electric Power System Optimization
- Sensorless Control of Electric Motors
- Electric Motor Design and Analysis
- Water Systems and Optimization
- Rough Sets and Fuzzy Logic
- Multilevel Inverters and Converters
- Simulation and Modeling Applications
- Multi-Criteria Decision Making
- Water-Energy-Food Nexus Studies
- Reservoir Engineering and Simulation Methods
- Advanced Decision-Making Techniques
- Hydrology and Watershed Management Studies
- Optimal Power Flow Distribution
- Geoscience and Mining Technology
- Civil and Geotechnical Engineering Research
- Advanced Computational Techniques and Applications
- Geotechnical Engineering and Underground Structures
- Advanced Algorithms and Applications
- Advanced Sensor and Control Systems
- Geotechnical Engineering and Soil Stabilization
- Data Mining Algorithms and Applications
Southwest Minzu University
2025
Zhengzhou University
2022-2024
Georgia Institute of Technology
2019-2023
Hohai University
2018-2021
Microsoft (United States)
2019-2020
Henan Normal University
2018-2020
Hunan Normal University
2020
UNSW Sydney
2020
Zhejiang University of Science and Technology
2016-2018
Shanghai Jiao Tong University
2018
Attribute reduction is one of the biggest challenges encountered in computational intelligence, data mining, pattern recognition, and machine learning. Effective feature selection as rough set theory is, it can only handle symbolic attributes. In order to overcome this drawback, fuzzy model proposed, which an extended sets able deal with imprecision uncertainty both numerical The existing attribute algorithms based on mainly take angle "attribute set," means they define object function...
In the last few years, an emerging trend in automatic speech recognition research is study of end-to-end (E2E) systems. Connectionist Temporal Classification (CTC), Attention Encoder-Decoder (AED), and RNN Transducer (RNN-T) are most popular three methods. Among these methods, RNN-T has advantages to do online streaming which challenging AED it doesn't have CTC's frame-independence assumption. this paper, we improve training two aspects. First, optimize algorithm reduce memory consumption so...
Attribute selection is considered as the most characteristic result in rough set theory to distinguish itself other theories. However, existing attribute approaches can not handle partially labeled data. So far, few studies on data have been conducted. In this paper, concept of discernibility pair based raised construct a uniform measure for attributes both supervised framework and unsupervised framework. Based pair, two kinds semisupervised algorithm are developed categorical Experiments...
Rough set theory, as one of the most useful soft computing methods dealing with vague and uncertain information, has been successfully applied to many fields, its main applications is perform attribute reduction. Although heuristic reduction algorithms have proposed within framework rough these are still computationally time consuming. In order overcome this deficit, we propose, in paper, two quick feature selection based on neighbor inconsistent pair, which can reduce consumed finding a...
Although great progress has been made in automatic speech recognition, significant performance degradation still exists noisy environments. Our previous work demonstrated the superior noise robustness of very deep convolutional neural networks (VDCNN). Based on our VDCNNs, this paper proposes a more advanced model referred to as residual network (VDCRN). This new incorporates batch normalization and learning, showing than VDCNNs.Then, alleviate mismatch between training testing conditions,...
Data augmentation is an effective method to increase the size of training data and reduce mismatch between testing for noise robust speech recognition. Different from traditional approaches by directly adding original waveform, in this work we utilize generative adversarial networks (GAN) generation improve recognition under conditions. With method, generated samples are based on spectrum feature level produced frame without dependence among them, augmented has no true labels. Then...
In this technical report, we present a joint effort of four groups, namely GT, USTC, Tencent, and UKE, to tackle Task 1 - Acoustic Scene Classification (ASC) in the DCASE 2020 Challenge. comprises two different sub-tasks: (i) 1a focuses on ASC audio signals recorded with multiple (real simulated) devices into ten fine-grained classes, (ii) 1b concerns classification data three higher-level classes using low-complexity solutions. For 1a, propose novel two-stage system leveraging upon ad-hoc...
To improve device robustness, a highly desirable key feature of competitive data-driven acoustic scene classification (ASC) system, novel two-stage system based on fully convolutional neural networks (CNNs) is proposed. Our leverages an ad-hoc score combination two CNN classifiers: (i) the first classifies inputs into one three broad classes, and (ii) second same ten finergrained classes. Three different architectures are explored to implement classifiers, frequency sub-sampling scheme...
The study on the electric property of conductive concrete is great importance to its application in engineering construction and other new fields as there no strict standard or specification formulated for testing conductivity concrete. This article, through preparation test specimen contrast methods concrete, studies process Through analysis research electrical author has obtained method mounting electrodes size voltage
Recently, the recurrent neural network transducer (RNN-T) architecture has become an emerging trend in end-to-end automatic speech recognition research due to its advantages of being capable for online streaming recognition. However, RNN-T training is made difficult by huge memory requirements, and complicated structure. A common solution ease employ connectionist temporal classification (CTC) model along with RNN language (RNNLM) initialize parameters. In this work, we conversely leverage...
Accents mismatching is a critical problem for end-to-end ASR. This paper aims to address this by building an accent-robust RNN-T system with domain adversarial training (DAT). We unveil the magic behind DAT and provide, first time, theoretical guarantee that learns accent-invariant representations. also prove performing gradient reversal in equivalent minimizing Jensen-Shannon divergence between output distributions. Motivated proof of equivalence, we introduce reDAT, novel technique based...
The Three Gorges cascade (TGC) reservoirs, China. Since uncertainties inherent in inflow forecasting, elevation-storage, discharge capacity, flood routing, and the preferences of decision makers (DMs) persist attribute measurements (AMs) weights (AWs), problems where multiple DMs select best compromise scheme for long-term comprehensive operation reservoirs considering numerous attributes become a stochastic multi-attribute group decision-making (MAGDM) process. To represent uncertain AMs...
The matrix completion problem is restoring a given with missing entries when handling incomplete data. In many existing researches, rank minimization plays central role in completion. this paper, noticing that the locally linear reconstruction can be used to approximate entries, we view from new perspective and propose an algorithm called approximation (LLA). LLA method tries keep local structure of data space while row angle column simultaneously. experimental results have demonstrated...
We propose a novel neural label embedding (NLE) scheme for the domain adaptation of deep network (DNN) acoustic model with unpaired data samples from source and target domains. With NLE method, we distill knowledge powerful source-domain DNN into dictionary embeddings, or l-vectors, one each senone class. Each l-vector is representation senone-specific output distributions learned to minimize average L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML"...
The river health evaluation is typically complex non-linear system with characteristics of fuzziness and randomness. However, conventional gray clustering method has difficult to effectively describe fuzzy random information simultaneously. For this purpose, the cloud model entropy theory are introduced establish 2D clustering-fuzzy comprehensive model. Different level models, it reflects situation from aspects corresponding water body complexity obtained by whitened weight function (first...