Puyang Zhao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Puyang Zhao has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Huiyun Zhang
Henan University
4 shared publications Gaigai Tang
2 shared publications- Zhiyi Yue2 shared publications
- Anne Buu
University of Texas Health Science Center at Houston, Houston, TX, USA
1 shared publication Guanghui Wang
1 shared publicationHeming Huang
1 shared publication
Latest Publications
- 4 June 2026
- Published by Elsevier in Applied Soft Computing
- Vol. 201
Residual attention network for audio-based speech emotion recognition
- 4 June 2026
- Published by Elsevier in Applied Soft Computing
- Vol. 201
Residual attention network for audio-based speech emotion recognition
- 4 December 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 114
Demography-Aware Personalized Federated Learning for fair, private, and efficient clinical risk prediction
- 4 December 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 114
Demography-Aware Personalized Federated Learning for fair, private, and efficient clinical risk prediction
- 7 October 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 112
Pre-attentive speech signal processing with adaptive routing for emotion recognition
- 7 October 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 112
Pre-attentive speech signal processing with adaptive routing for emotion recognition
- Top Cited Papers
- 8 May 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 109
Reproducible and generalizable speech emotion recognition via an Intelligent Fusion Network
- 8 May 2025
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 109
Top Cited PapersReproducible and generalizable speech emotion recognition via an Intelligent Fusion Network
- Top Cited Papers
- 10 February 2025
- Published by Taylor & Francis in The American Journal of Drug and Alcohol Abuse
- Vol. 51 (2) , 165-172
Applied statistical methods for identifying features of heart rate that are associated with nicotine vaping
Background: Wearable devices have been increasingly adopted to collect physiological data such as heart rate that may infer momentary risk of substance use. Yet, innovative methods capable for handling these complex time series data as presented in the statistics or data science literature may not be accessible to substance use researchers.Objectives: This study introduces a series of statistical methods to analyze heart rate data and identify features that are associated with nicotine vaping.Methods: Nontechnical description of the methods coupled with the information about open-source software packages that implemented these methods was provided. The analytical procedure included 5 steps: (1) de-noising by the singular spectrum analysis (SSA); (2) sleep region identification by the Sum of Single Effects (SuSiE) model; (3) repeated heart rate pattern identification by the matrix profile; (4) dimension reduction by the linear regression; and (5) comparing repeated heart rate patterns across non-vaping and vaping regions by the linear mixed model. Secondary analysis was conducted on heart rate and ecological momentary assessment (EMA) data collected from 35 young adult e-cigarette users (66% female) for 7 days.Results: Effectiveness of the methods was demonstrated by graphical presentations showing that the extracted features characterize sleep patterns and heart rate changes before and after vaping events quite well. Secondary analysis found that heart rate was higher and changed faster before vaping.Conclusion: Statistical methods can effectively extract useful features from heart rate data that may inform momentary vaping risk and optimal timings for delivering messages in mobile-phone based interventions.
- 10 February 2025
- Published by Taylor & Francis in The American Journal of Drug and Alcohol Abuse
- Vol. 51 (2) , 165-172
Top Cited PapersApplied statistical methods for identifying features of heart rate that are associated with nicotine vaping
Background: Wearable devices have been increasingly adopted to collect physiological data such as heart rate that may infer momentary risk of substance use. Yet, innovative methods capable for handling these complex time series data as presented in the statistics or data science literature may not be accessible to substance use researchers.Objectives: This study introduces a series of statistical methods to analyze heart rate data and identify features that are associated with nicotine vaping.Methods: Nontechnical description of the methods coupled with the information about open-source software packages that implemented these methods was provided. The analytical procedure included 5 steps: (1) de-noising by the singular spectrum analysis (SSA); (2) sleep region identification by the Sum of Single Effects (SuSiE) model; (3) repeated heart rate pattern identification by the matrix profile; (4) dimension reduction by the linear regression; and (5) comparing repeated heart rate patterns across non-vaping and vaping regions by the linear mixed model. Secondary analysis was conducted on heart rate and ecological momentary assessment (EMA) data collected from 35 young adult e-cigarette users (66% female) for 7 days.Results: Effectiveness of the methods was demonstrated by graphical presentations showing that the extracted features characterize sleep patterns and heart rate changes before and after vaping events quite well. Secondary analysis found that heart rate was higher and changed faster before vaping.Conclusion: Statistical methods can effectively extract useful features from heart rate data that may inform momentary vaping risk and optimal timings for delivering messages in mobile-phone based interventions.
- Top Cited Papers
- 31 January 2025
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 144
Sparse temporal aware capsule network for robust speech emotion recognition
- 31 January 2025
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 144
Top Cited PapersSparse temporal aware capsule network for robust speech emotion recognition
- Open Access
An Attention-based Long Short-Term Memory Framework for Detection of Bitcoin Scams
Bitcoin is the most common cryptocurrency involved in cyber scams. Cybercriminals often utilize pseudonymity and privacy protection mechanism associated with Bitcoin transactions to make their scams virtually untraceable. The Ponzi scheme has attracted particularly significant attention among Bitcoin fraudulent activities. This paper considers a multi-class classification problem to determine whether a transaction is involved in Ponzi schemes or other cyber scams, or is a non-scam transaction. We design a specifically designed crawler to collect data and propose a novel Attention-based Long Short-Term Memory (A-LSTM) method for the classification problem. The experimental results show that the proposed model has better efficiency and accuracy than existing approaches, including Random Forest, Extra Trees, Gradient Boosting, and classical LSTM. With correctly identified scam features, our proposed A-LSTM achieves an F1-score over 82% for the original data and outperforms the existing approaches.
Open AccessAn Attention-based Long Short-Term Memory Framework for Detection of Bitcoin Scams
Bitcoin is the most common cryptocurrency involved in cyber scams. Cybercriminals often utilize pseudonymity and privacy protection mechanism associated with Bitcoin transactions to make their scams virtually untraceable. The Ponzi scheme has attracted particularly significant attention among Bitcoin fraudulent activities. This paper considers a multi-class classification problem to determine whether a transaction is involved in Ponzi schemes or other cyber scams, or is a non-scam transaction. We design a specifically designed crawler to collect data and propose a novel Attention-based Long Short-Term Memory (A-LSTM) method for the classification problem. The experimental results show that the proposed model has better efficiency and accuracy than existing approaches, including Random Forest, Extra Trees, Gradient Boosting, and classical LSTM. With correctly identified scam features, our proposed A-LSTM achieves an F1-score over 82% for the original data and outperforms the existing approaches.