Yao Ying
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yao Ying has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Peng Shan
College of Information Science and Engineering, Northeastern University, Shenyang, China
2 shared publications - Qiaoyun Wang
School of Chemistry University of New South Wales Sydney NSW 2052 Australia
2 shared publications Silong Peng
Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China
2 shared publications- Silong Peng
Institute of Automation, Chinese Academy of Sciences, Beijing, China
2 shared publications - Yuhui Zhao
School of Management Northeastern University at Qinhuangdao, Qin Huang dao, Hebei, China
2 shared publications - Chengwen Luo
Shenzhen University, China
1 shared publication
Latest Publications
- Open Access
- 11 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 62-69
Design and Implementation of STM32 Embedded System
- 11 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 62-69
Open AccessDesign and Implementation of STM32 Embedded System
- 1 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 9 (18) , 18015-18027
Wi-Phrase: Deep Residual-Multihead Model for WiFi Sign Language Phrase Recognition
Sign language (SL) is used by hearing impaired and deaf people. The WiFi-based sign language recognition (SLR) technology has attracted much attention due to its contactless nature and wide applications. Most previous SLR works are designed to recognize a single isolated sign word in data samples. However, such assumption is not realistic in practical applications, such as, in daily communication, deaf people usually express their mind through a phrase (a.k.a. a group of words) rather than an isolated word. Compared with the previous works, the sign words in a phrase have variety of length, sequential patterns, and combinations in realistic communications between deaf people. Therefore, it is challenging in exploiting the WiFi signals to accurately capture the unique patterns of SL in the previous natural setting. In this article, we propose Wi-Phrase, a multigesture context-awareness SLR system. Wi-Phrase exploits WiFi signals to translate SL to the English phrase. To achieve this, Wi-Phrase employs principal component analysis (PCA) projection to filter out the noise and convert cleaned WiFi signals to spectrogram. Then, we propose a novel Residual-MultiHead model that exploit residual learn structure to obtain local patterns of phrases and adopt multihead block to capture the global context information of phrase. To prove the advanced nature of our model, we design a WiFi-based SL phrase data set of 40 categories for experiments. Our comprehensive evaluation shows that Wi-Phrase achieves an accurate phrase recognition accuracy of 95.03%. In future, we envision Wi-Phrase could be widely used as the phrase command control system for deaf people in IoT devices.
- 1 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 9 (18) , 18015-18027
Wi-Phrase: Deep Residual-Multihead Model for WiFi Sign Language Phrase Recognition
Sign language (SL) is used by hearing impaired and deaf people. The WiFi-based sign language recognition (SLR) technology has attracted much attention due to its contactless nature and wide applications. Most previous SLR works are designed to recognize a single isolated sign word in data samples. However, such assumption is not realistic in practical applications, such as, in daily communication, deaf people usually express their mind through a phrase (a.k.a. a group of words) rather than an isolated word. Compared with the previous works, the sign words in a phrase have variety of length, sequential patterns, and combinations in realistic communications between deaf people. Therefore, it is challenging in exploiting the WiFi signals to accurately capture the unique patterns of SL in the previous natural setting. In this article, we propose Wi-Phrase, a multigesture context-awareness SLR system. Wi-Phrase exploits WiFi signals to translate SL to the English phrase. To achieve this, Wi-Phrase employs principal component analysis (PCA) projection to filter out the noise and convert cleaned WiFi signals to spectrogram. Then, we propose a novel Residual-MultiHead model that exploit residual learn structure to obtain local patterns of phrases and adopt multihead block to capture the global context information of phrase. To prove the advanced nature of our model, we design a WiFi-based SL phrase data set of 40 categories for experiments. Our comprehensive evaluation shows that Wi-Phrase achieves an accurate phrase recognition accuracy of 95.03%. In future, we envision Wi-Phrase could be widely used as the phrase command control system for deaf people in IoT devices.
- RetractedOpen Access
- 12 May 2021
- Published by Wiley in Security and Communication Networks
- Vol. 2021, 1-9
Improving Convolutional Neural Networks with Competitive Activation Function
The activation function is the basic component of the convolutional neural network (CNN), which provides the nonlinear transformation capability required by the network. Many activation functions make the original input compete with different linear or nonlinear mapping terms to obtain different nonlinear transformation capabilities. Until recently, the original input of funnel activation (FReLU) competed with the spatial conditions, so FReLU not only has the ability of nonlinear transformation but also has the ability of pixelwise modeling. We summarize the competition mechanism in the activation function and then propose a novel activation function design template: competitive activation function (CAF), which promotes competition among different elements. CAF generalizes all activation functions that use competition mechanisms. According to CAF, we propose a parametric funnel rectified exponential unit (PFREU). PFREU promotes competition among linear mapping, nonlinear mapping, and spatial conditions. We conduct experiments on four datasets of different sizes, and the experimental results of three classical convolutional neural networks proved the superiority of our method.
- 12 May 2021
- Published by Wiley in Security and Communication Networks
- Vol. 2021, 1-9
RetractedOpen AccessImproving Convolutional Neural Networks with Competitive Activation Function
The activation function is the basic component of the convolutional neural network (CNN), which provides the nonlinear transformation capability required by the network. Many activation functions make the original input compete with different linear or nonlinear mapping terms to obtain different nonlinear transformation capabilities. Until recently, the original input of funnel activation (FReLU) competed with the spatial conditions, so FReLU not only has the ability of nonlinear transformation but also has the ability of pixelwise modeling. We summarize the competition mechanism in the activation function and then propose a novel activation function design template: competitive activation function (CAF), which promotes competition among different elements. CAF generalizes all activation functions that use competition mechanisms. According to CAF, we propose a parametric funnel rectified exponential unit (PFREU). PFREU promotes competition among linear mapping, nonlinear mapping, and spatial conditions. We conduct experiments on four datasets of different sizes, and the experimental results of three classical convolutional neural networks proved the superiority of our method.
- 1 March 2021
- Published by Elsevier in Chemometrics and Intelligent Laboratory Systems
- Vol. 210, 104247
A nonlinear calibration transfer method based on joint kernel subspace
- 1 March 2021
- Published by Elsevier in Chemometrics and Intelligent Laboratory Systems
- Vol. 210, 104247
A nonlinear calibration transfer method based on joint kernel subspace
- Open Access
- 15 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 101633-101640
Rectified Exponential Units for Convolutional Neural Networks
Rectified linear unit (ReLU) plays an important role in today's convolutional neural networks (CNNs). In this paper, we propose a novel activation function called Rectified Exponential Unit (REU). Inspired by two recently proposed activation functions: Exponential Linear Unit (ELU) and Swish, the REU is designed by introducing the advantage of flexible exponent and multiplication function form. Moreover, we propose the Parametric REU (PREU) to increase the expressive power of the REU. The experiments with three classical CNN architectures, LeNet-5, Network in Network, and Residual Network (ResNet) on scale-various benchmarks including Fashion-MNIST, CIFAR10, CIFAR100, and Tiny ImageNet demonstrate that REU and PREU achieve improvement compared with other activation functions. Our results show that our REU has relative error improvements over ReLU of 7.74% and 6.08% on CIFAR-10 and 100 with the ResNet, the improvements of PREU is 9.24% and 9.32%. Finally, we use the different PREU variants in the Residual unit to achieve more stable results.
- 15 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 101633-101640
Open AccessRectified Exponential Units for Convolutional Neural Networks
Rectified linear unit (ReLU) plays an important role in today's convolutional neural networks (CNNs). In this paper, we propose a novel activation function called Rectified Exponential Unit (REU). Inspired by two recently proposed activation functions: Exponential Linear Unit (ELU) and Swish, the REU is designed by introducing the advantage of flexible exponent and multiplication function form. Moreover, we propose the Parametric REU (PREU) to increase the expressive power of the REU. The experiments with three classical CNN architectures, LeNet-5, Network in Network, and Residual Network (ResNet) on scale-various benchmarks including Fashion-MNIST, CIFAR10, CIFAR100, and Tiny ImageNet demonstrate that REU and PREU achieve improvement compared with other activation functions. Our results show that our REU has relative error improvements over ReLU of 7.74% and 6.08% on CIFAR-10 and 100 with the ResNet, the improvements of PREU is 9.24% and 9.32%. Finally, we use the different PREU variants in the Residual unit to achieve more stable results.