Proposed blind reverberation time estimation using deep neural networks (DNN)
using multi-channel microphone
Proposed both Single channel & multi-channel methods
- Obtaining the acoustic parameter from the sound source
- Conducted an intensive study with single & multiple channel based algorithm using neural networks.
(Published in IEEE and Acta Acustica)
- Developed estimation algorithm for dereverberation and an acoustic model (C, MATLAB)
- Contributed distributive research with LG electronics
◇ Myungin Lee, Joon-Hyuk Chang, "Deep neural network based blind estimation of reverberation time based on multi-channel microphones," Acta Acustica united with Acustica, 2018.
◇ Myungin Lee, Joon-Hyuk Chang, “Blind Estimation of Reverberation Time on Multi-Channel Microphone using Deep Neural Network,” Master’s thesis, Feb., 2017.
◇ Myungin Lee, Joon-Hyuk Chang, “Blind Estimation of Reverberation Time using Deep Neural Network,” IEEE International Conference on Network Infrastructure and Digital Content (IC-NIDC), Beijing, China, Sep., 2016.
- Multichannel Microphone-based Reverberation Time Estimation Method and Device which use Deep Neural Network Technical Field, WO/US Patent, 2017.
- Multi-Channel Microphone based Reverberation Time Estimation using Deep Neural Network, Korea Patent, 2016.
Performance of T60 estimation algorithms
in various noise environments for different SNRs:
(a) bias (b) MSE, and (c) ρ.