Penglong Wu

PLA Army Engineering University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Xia Zou

PLA Army Engineering University

Meng Sun

PLA Army Engineering University

Xingyu Zhang

Harbin Institute of Technology

Jun He

National University of Defense Technology

Yi‐Min Wang

Qingdao University of Science and Technology

Xingyu Zhang

North China Electric Power University

Xingyu Zhang

University of Pittsburgh Medical Center

Li Li

Zaozhuang University

Cooperative Institutions

PLA Army Engineering University

Harbin Institute of Technology

University of Hong Kong

Xidian University

Tsinghua University

Huazhong University of Science and Technology

Qingdao University of Science and Technology

National University of Singapore

Sun Yat-sen University

South China University of Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Robust Speaker Recognition Using Improved GFCC and Adaptive Feature Selection

Advances in intelligent systems and computing (2019)

Xingyu Zhang Xia Zou Meng Sun Penglong Wu

Mel-frequency cepstrum

Robustness

Normalization

10.1007/978-3-030-16946-6_13

Cite

Citations (3)

On the Complementary Role of DNN Multi-Level Enhancement for Noisy Robust Speaker Recognition in an I-Vector Framework

IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences (2019)

Xingyu Zhang Xia Zou Meng Sun Penglong Wu Yi‐Min Wang

In order to improve the noise robustness of automatic speaker recognition, many techniques on speech/feature enhancement have been explored by using deep neural networks (DNN). In this work, a DNN multi-level enhancement (DNN-ME), which consists of the stages of signal enhancement, cepstrum enhancement and i-vector enhancement, is proposed for text-independent speaker recognition. Given the fact that these enhancement methods are applied in different stages of the speaker recognition pipeline, it is worth exploring the complementary role of these methods, which benefits the understanding of the pros and cons of the enhancements of different stages. In order to use the capabilities of DNN-ME as much as possible, two kinds of methods called Cascaded DNN-ME and joint input of DNNs are studied. Weighted Gaussian mixture models (WGMMs) proposed in our previous work is also applied to further improve the model's performance. Experiments conducted on the Speakers in the Wild (SITW) database have shown that DNN-ME demonstrated significant superiority over the systems with only a single enhancement for noise robust speaker recognition. Compared with the i-vector baseline, the equal error rate (EER) was reduced from 5.75 to 4.01.

10.1587/transfun.2019eal2104

Cite

Citations (1)

Adaptive Consistent Dictionary Learning for Audio Declipping

Lecture notes in electrical engineering (2019)

Penglong Wu Xia Zou Meng Sun Li Li Xingyu Zhang

Clipping (morphology)

Distortion (music)

Degree (music)

10.1007/978-981-13-8707-4_7

Cite

Citations (0)

The Influence of Clipping on the Performance of a Low Bit Rate Parametric Speech Coder

2021 14th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI) (2019)

Penglong Wu Xia Zou Meng Sun Jun He Xingyu Zhang

This paper presents a study of the influence of clipping on low bit rate parametric speech coder. By selecting three parameters of MELP vocoder: the fundamental frequency, voicing and LPC, the influence of different clipping degree on the MELP vocoder is studied. In the experimental stage, the performance of the MELP vocoder at different clipping degree are evaluated by using pitch estimation accuracy rate, voicing decision error rate and average Log-spectral distance. And then, the reasons which lead to a decrease in the performance of parameter extraction are analyzed. The experimental results show that clipping has a certain influence on the accuracy of pitch estimation and voicing decision. However, the accuracy of pitch estimation and voicing decision can keep high accuracy even when the clipping degree is large, which indicates that the pitch and voicing are strongly resistant to clipping. For LPC, the increase of the clipping degree increases the average Log-spectral distance, resulting in an inaccurate estimation of LPC.

Clipping (morphology)

10.1109/cisp-bmei48845.2019.8965790

Cite

Citations (0)