A Joint Learning Algorithm for Complex-Valued T-F Masks in Deep Learning-Based Single-Channel Speech Enhancement Systems

2019 ◽  
Vol 27 (6) ◽  
pp. 1098-1108 ◽  
Author(s):  
Jinkyu Lee ◽  
Hong-Goo Kang
Author(s):  
Yantao Chen ◽  
Binhong Dong ◽  
Xiaoxue Zhang ◽  
Pengyu Gao ◽  
Shaoqian Li

IEEE Access ◽  
2021 ◽  
Vol 9 ◽  
pp. 64524-64538
Author(s):  
Sujan Kumar Roy ◽  
Aaron Nicolson ◽  
Kuldip K. Paliwal

2021 ◽  
Author(s):  
Chunliu Shi

Abstract In order to improve the effect of intelligent language translation, this paper analyzes the problems of the MSE cost function used by most of the current DNN-based speech enhancement algorithms, and proposes a deep learning speech enhancement algorithm based on perception-related cost functions. Moreover, this paper embeds the suppression gain parameter estimation into the architecture of the traditional speech enhancement algorithm, and converts the relationship between the noisy speech spectrum and the enhanced speech spectrum into a simple multiplication relationship based on suppression gain combined with deep learning algorithms to construct an intelligent language translation system. Moreover, this paper evaluates the translation effect of the system, analyzes the actual results, and uses simulation tests to verify the performance of the intelligent language translation model constructed in this paper. From the experimental results, it can be seen that the intelligent language translation system based on deep learning algorithms has good results.


2020 ◽  
Vol 8 (6) ◽  
pp. 4788-4794

Today’s young generation is tormented by anxiety and stress. A past study shows that, anxiety and stress negatively impact mental and physical health, which ultimately ends up in loss of confidence, self-esteem, and negative performance. This work set guidelines for a replacement approach in neuroscience that only single-channel EEG data has sufficient information for emotion recognition. In this paper, the performance of system is evaluated using subject independent test on the SEED benchmark database using deep learning algorithm namely a bidirectional long short term memory (BiLSTM) classifier. The performance shows that results of single-channel FP1, the combined band (theta, alpha, beta, and gamma) are similar to 62 channels the best accuracy of the beta band. Result obtained for single channel (FP1) using differential entropy (DE) for all band is 74.91% as that of the highest accuracy of the beta band for 62 channels Yang Li, W. Zheng, 2019 74.85%.


2018 ◽  
Vol 25 (8) ◽  
pp. 1276-1280 ◽  
Author(s):  
Jinkyu Lee ◽  
Jan Skoglund ◽  
Turaj Shabestary ◽  
Hong-Goo Kang

Sign in / Sign up

Export Citation Format

Share Document