A5100681256- Speech and Audio Processing
- Speech Recognition and Synthesis
- Financial Markets and Investment Strategies
- Corporate Social Responsibility Reporting
- Image Processing Techniques and Applications
- Music and Audio Processing
- EEG and Brain-Computer Interfaces
- Advanced Vision and Imaging
- Blind Source Separation Techniques
- Environmental Sustainability in Business
- Advanced Neural Network Applications
- Statistical Methods and Inference
- Optical measurement and interference techniques
- Housing Market and Economics
- Stock Market Forecasting Methods
- Web Application Security Vulnerabilities
- Energy, Environment, and Transportation Policies
- Text and Document Classification Technologies
- Remote-Sensing Image Classification
- Multimodal Machine Learning Applications
- Mathematical Analysis and Transform Methods
- Monetary Policy and Economic Impact
- Housing, Finance, and Neoliberalism
- Phonetics and Phonology Research
- Image Enhancement Techniques
Zhoukou Normal University
2025
Peking University
2024
Beijing Institute of Technology
2019-2024
Australian National University
2020-2021
Data61
2021
Commonwealth Scientific and Industrial Research Organisation
2021
Tsinghua University
2017-2019
University of Electronic Science and Technology of China
2017-2019
Radboud University Nijmegen
2005-2007
This paper investigated the impact of employee quality on corporate social responsibility (CSR). Based data from China A-share-listed companies for years 2012–2016 and using ordinary least squares, our empirical results show that educational level workforce, as a proxy quality, is positively associated with CSR, which suggests higher education can promote CSR implementation. Additional analyses found this positive relationship more pronounced in non-state-owned enterprises, enterprises...
In this paper we first demonstrate continuous noisy speech recognition using electroencephalography (EEG) signals on English vocabulary different types of state the art end-to-end automatic (ASR) models, further provide results obtained EEG data recorded under experimental conditions. We finally decoding spectrum from a long short term memory (LSTM) based regression model and Generative Adversarial Network (GAN) model. Our feasibility for conditions preliminary synthesis features.
The purpose of this research is to consider if the growing popularity sustainable investment does not create additional risks in investing. Different views on investments were analyzed identify different approaches main risks. A quantitative analysis was carried out investigate possible benefits and advantages investment. Without taking into account social perks investing funds, study evaluates performance economic returns both traditional funds. two parts by comparing samples 30 Firstly,...
Dairy goats, a livestock species with long history of milk production, are essential for the economic advancement nations, particularly in regions experiencing growth. In this study, we gathered whole-genome resequencing data 58 including 34 dairy goats and 24 wild (Bezoar), to explore selection signatures linked production traits using ROH (Runs homozygosity), CLR (composite likelihood ratio), Fst (Fixation index), XP-EHH (Ex-tended haplotype homozygosity across populations)...
Along with the rapid development of digital information technology, e-government is great potential because it a new form conducting public administration and way demonstrating governmental innovation. The literature suggests that foreign direct investment (FDI) increasingly associated continuing in China. Using Annual Census Industrial Enterprises scores government portals, this study examines effects on FDI how subsidies mediate relationships between FDI. Our results show positively...
Estimating scene depth from a single image can be widely applied to understand 3D environments due the easy access of images captured by consumer-level cameras. Previous works exploit conditional random fields (CRFs) estimate depth, where neighboring pixels (superpixels) with similar appearances are constrained share same depth. However, may vary significantly in slanted surface, thus leading severe estimation errors. In order eliminate those errors, we propose superpixel-based normal guided...
Investors’ beliefs are the driving force behind trading of stocks and, hence, sustainable stock returns. Although investors’ usually unobservable, this study develops a new approach to estimate investors beliefs. Following well-established rational learning and market microstructure models, it is assumed that informed traders submit orders according their beliefs, whereas makers/uninformed make Bayesian inferences about traders’ private signals after observing total order flows. By fitting...
Studying the driving factors of environmental pollution is great importance for China. Previous literature mainly focused on cause national aggregate emission changes. However, research about effect fiscal expenditures science and technology (FESTs) rare. Considering large gap among cities in China, it necessary to investigate whether how FESTs affect cities. We adopted three kinds typical pollutants including sulfur dioxide (SO2) emissions, wastewater emission, atmospheric particulate...
Noise is ubiquitous in the world around us. Difficulty estimating noise within a dataset makes learning from such difficult and challenging task. In this paper, we propose novel effective framework order to alleviate adverse effects of dataset. Towards aim, modify collaborative training utilize discrepancy constraints between respective feature extractors enabling distinct, yet discriminative features, pacifying noise. Empirical results our proposed algorithm, Discrepant Collaborative...
In this paper we introduce a novel method for clustering speech gestures, represented as contin uous trajectories in acoustic parameter space.Trajectory Clustering allows us to avoid the conditional independence assumption that makes it difficult account fact successive measurements of an articulatory gesture are correlated.We apply Trajectory developing multiple parallel HMMs continuous digits recognition task.We compare performance obtained with data-driven conventional Head-Body-Tail...
The focal length information of an image is indispensable for many computer vision tasks. In general, can be obtained via camera calibration using specific planner patterns. However, images taken by unknown device, only estimated based on the itself. Currently, most single-image estimation methods make use predefined geometric cues (such as vanishing points or parallel lines) to infer length, which constrains their applications mainly manmade scenes. machine learning algorithms have...
We undertook an autopsy of the drivers individual foreign real estate investment ‘bust’ in Australia through a new theoretical lens ‘habitus’. Our data drew contours around capital ‘boom and bust’ cycle, as well long-term commitment professionals sector to Australia’s market. More specifically, we showed that cycle began earnest about 2010 (starting at A$8.7 billion), grew A$72.4 billion 2016–2017, then declined A$12.5 2017–2018. This decline into Australian occurred within domestic market...
Recent research suggests that modeling coarticulation in speech is more appropriate at the syllable level. However, due to a number of additional factors affect way syllables are articulated, creating multiple paths through models might be necessary. Our previous on longer-length multi-path connected digit recognition has proved trajectory clustering an attractive approach deriving models. In this paper, we extend our large vocabulary continuous by clusters for 94 very frequent 20-hour data...
Abstract Building on the Bayesian Theorem, we propose a multi‐period market microstructure model to understand how investors underact new information and duration of underreaction. Applying post‐earnings‐announcement drifts, our simulation regression analyses show that post‐announcement price adjustment process drifts can be explained by measure belief updating speed quantifies uncertainties faced when incorporating into prices. Our study highlights importance uninformed in explaining...
Context dependent modelling is known to improve recognition performance for automatic speech recognition. One of the major limitations, especially approaches based on Decision Trees, that questions thatguidethesearchfor effectivecontextsmustbeknown in advance. However, variation signals caused by multiple factors, not all which may be during training procedure. State tying methods, other hand, are strictly local, and therefore do allow reap benefits spans longer length units such as...
Segmented regression models offer model flexibility and interpretability as compared to the global parametric nonparametric models, yet are challenging in both estimation inference. We consider a four-regime segmented for temporally dependent data with segmenting boundaries depending on multivariate covariates non-diminishing boundary effects. A mixed integer quadratic programming algorithm is formulated facilitate least square of parameters. The rates convergence asymptotic distributions...