Yue Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yue Zhang has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
Jinxia Yu
School of Computer Science and Technology, Henan Polytechnic University, Jiaozuo, China
2 shared publications- Luyao Liu
Department of Software, Henan Polytechnic University, Jiaozuo 454150, China
2 shared publications - Wei Zhao
Department of Software, Henan Polytechnic University, Jiaozuo 454150, China
2 shared publications - Wenzhe Liu
Department of Information Engineering, Huzhou University, Huzhou 313000, China
2 shared publications Yongli Tang
School of Software, Henan Polytechnic University, Jiaozuo, China
2 shared publicationsZhenlun Gao
School of Software, Henan Polytechnic University, Jiaozuo, China
2 shared publications
Latest Publications
- Open Access
- 15 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (20) , 4741
LoRA-Efficient Multi-Task Learning with Joint Semantic and Paralinguistic Distillation for Speech Emotion Recognition
Speech Emotion Recognition (SER) is an important but challenging research topic in affective computing, aiming to identify the emotional state of a speaker from their speech. Emotional information in an utterance is carried by two complementary channels: the semantic content of the words (what is said) and the paralinguistic manner of delivery (how it is said). Self-supervised speech models such as WavLM capture acoustic patterns well, but they do not explicitly model lexical-semantic emotional cues. Multimodal alternatives, on the other hand, depend on automatic speech recognition (ASR) transcripts at inference time, increasing the deployment cost. To address these issues, we propose a LoRA-efficient multi-task learning framework built on a WavLM backbone, in which the semantic and paralinguistic representations of emotional speech are learned jointly inside a single speech encoder. The framework trains three objectives together: semantic distillation, paralinguistic self-reconstruction, and emotion classification, where the latter is treated as the primary task and the other two provide supporting supervision. Experiments on MSP-IMPROV and IEMOCAP validate the effectiveness of the proposed framework, and on IEMOCAP the model outperforms several existing SER methods, confirming the rationality of the design.
- 15 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (20) , 4741
Open AccessLoRA-Efficient Multi-Task Learning with Joint Semantic and Paralinguistic Distillation for Speech Emotion Recognition
Speech Emotion Recognition (SER) is an important but challenging research topic in affective computing, aiming to identify the emotional state of a speaker from their speech. Emotional information in an utterance is carried by two complementary channels: the semantic content of the words (what is said) and the paralinguistic manner of delivery (how it is said). Self-supervised speech models such as WavLM capture acoustic patterns well, but they do not explicitly model lexical-semantic emotional cues. Multimodal alternatives, on the other hand, depend on automatic speech recognition (ASR) transcripts at inference time, increasing the deployment cost. To address these issues, we propose a LoRA-efficient multi-task learning framework built on a WavLM backbone, in which the semantic and paralinguistic representations of emotional speech are learned jointly inside a single speech encoder. The framework trains three objectives together: semantic distillation, paralinguistic self-reconstruction, and emotion classification, where the latter is treated as the primary task and the other two provide supporting supervision. Experiments on MSP-IMPROV and IEMOCAP validate the effectiveness of the proposed framework, and on IEMOCAP the model outperforms several existing SER methods, confirming the rationality of the design.
- Open Access
- 12 June 2026
- Published by MDPI AG in Electronics
- Vol. 15 (12) , 2608
Not All Segments Are Needed: Lightweight Adaptive Pre-Selection of Emotional Speech Segments
The extraction of speech emotion features—particularly at the utterance level—constitutes a critical yet challenging aspect of Speech Emotion Recognition (SER). Emotion represents high-level paralinguistic information, and not all segments within a speech signal carry emotionally salient cues, especially in longer utterances. Conventional methods that process entire utterances may thus waste computational resources and introduce irrelevant acoustic interference. To address this, we propose AdaPre-Selection, an adaptive pre-selection mechanism designed to identify and extract emotion-rich segments from speech signals. Acting as a flexible front-end compression module, AdaPre-Selection consists of two complementary components: an Active Emotion Positioning (AEP) module and a Passive Emotion Constraint (PEC) module. Within AEP, Temporal Length Selection (TLS) and Start Time Point Selection (STPS) operate jointly to adaptively locate the optimal emotional segment in each utterance. Evaluated on two benchmark datasets (IEMOCAP and MSP-IMPROV) using four state-of-the-art SER models, AdaPre-Selection consistently outperforms common preprocessing baselines and delivers the most substantial improvement in recognition performance.
- 12 June 2026
- Published by MDPI AG in Electronics
- Vol. 15 (12) , 2608
Open AccessNot All Segments Are Needed: Lightweight Adaptive Pre-Selection of Emotional Speech Segments
The extraction of speech emotion features—particularly at the utterance level—constitutes a critical yet challenging aspect of Speech Emotion Recognition (SER). Emotion represents high-level paralinguistic information, and not all segments within a speech signal carry emotionally salient cues, especially in longer utterances. Conventional methods that process entire utterances may thus waste computational resources and introduce irrelevant acoustic interference. To address this, we propose AdaPre-Selection, an adaptive pre-selection mechanism designed to identify and extract emotion-rich segments from speech signals. Acting as a flexible front-end compression module, AdaPre-Selection consists of two complementary components: an Active Emotion Positioning (AEP) module and a Passive Emotion Constraint (PEC) module. Within AEP, Temporal Length Selection (TLS) and Start Time Point Selection (STPS) operate jointly to adaptively locate the optimal emotional segment in each utterance. Evaluated on two benchmark datasets (IEMOCAP and MSP-IMPROV) using four state-of-the-art SER models, AdaPre-Selection consistently outperforms common preprocessing baselines and delivers the most substantial improvement in recognition performance.
- 13 February 2026
- Published by Springer Nature in Quantum Information Processing
- Vol. 25 (2) , 66
Qnnformer: quantum-enhanced neural networks with multi-attention mechanisms for long-term time series forecasting
- 13 February 2026
- Published by Springer Nature in Quantum Information Processing
- Vol. 25 (2) , 66
Qnnformer: quantum-enhanced neural networks with multi-attention mechanisms for long-term time series forecasting
- Open Access
- 14 November 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 39969
Quantum-enhanced dual-layer graph attention network for time-series forecasting
Time-series forecasting plays a pivotal role in domains such as traffic prediction, financial analysis, and energy consumption monitoring. However, real-world time series often exhibit intertwined patterns–trends, seasonality, and latent structures–that pose significant challenges to forecasting accuracy. This paper proposes a novel forecasting model named QFreqFormer, which stands for Quantum Frequency Transformer. It combines the Quantum Fourier Transform (QFT) with a Dual-Layer Graph Attention Network (D-PAD) to effectively tackle the complexities of time-series forecasting. The QFT module exploits quantum parallelism and superposition to decompose time-series data into frequency components, offering a compact spectral representation. To further enhance the model’s ability to capture intricate multi-frequency patterns, we propose the Quantum Frequency Decomposition-Reconstruction (Q-FR-Q) module, which progressively separates high- and low-frequency components using quantum parallel processing. The D-PAD framework integrates Graph Convolutional Networks (GCNs) with attention mechanisms to dynamically model temporal dependencies across frequency layers. Experimental results on benchmark datasets demonstrate that the proposed model consistently outperforms state-of-the-art methods in terms of Mean Squared Error (MSE) and Mean Absolute Error (MAE) across various horizons. In addition, the model demonstrates strong transfer learning capability, underscoring its robustness and generalizability across heterogeneous forecasting scenarios. This study introduces a quantum-enhanced deep learning framework that improves both forecasting accuracy and computational efficiency, offering practical advantages in real-world applications.
- 14 November 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 39969
Open AccessQuantum-enhanced dual-layer graph attention network for time-series forecasting
Time-series forecasting plays a pivotal role in domains such as traffic prediction, financial analysis, and energy consumption monitoring. However, real-world time series often exhibit intertwined patterns–trends, seasonality, and latent structures–that pose significant challenges to forecasting accuracy. This paper proposes a novel forecasting model named QFreqFormer, which stands for Quantum Frequency Transformer. It combines the Quantum Fourier Transform (QFT) with a Dual-Layer Graph Attention Network (D-PAD) to effectively tackle the complexities of time-series forecasting. The QFT module exploits quantum parallelism and superposition to decompose time-series data into frequency components, offering a compact spectral representation. To further enhance the model’s ability to capture intricate multi-frequency patterns, we propose the Quantum Frequency Decomposition-Reconstruction (Q-FR-Q) module, which progressively separates high- and low-frequency components using quantum parallel processing. The D-PAD framework integrates Graph Convolutional Networks (GCNs) with attention mechanisms to dynamically model temporal dependencies across frequency layers. Experimental results on benchmark datasets demonstrate that the proposed model consistently outperforms state-of-the-art methods in terms of Mean Squared Error (MSE) and Mean Absolute Error (MAE) across various horizons. In addition, the model demonstrates strong transfer learning capability, underscoring its robustness and generalizability across heterogeneous forecasting scenarios. This study introduces a quantum-enhanced deep learning framework that improves both forecasting accuracy and computational efficiency, offering practical advantages in real-world applications.