Audio samples
All samples have been rescaled to –26 dBov based on the active speech level (ASL).
Reference:
@article{song2022improvedcem,
title={Improved CEM for speech harmonic enhancement in single channel noise suppression},
author={Song, Yanjue and Madhu, Nilesh},
journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},
volume={30},
number={},
pages={2492-2503},
year={2022},
publisher={IEEE},
doi={10.1109/TASLP.2022.3190725}
}
Methods |
Noisy input |
Baseline 1: LSA |
Baseline 2: CEMID |
Residual Amplitude Estimation |
Cepstrum Convolution |
Combined |
Clean Ref |