Blind Room Acoustics Characterization Using Recurrent Neural Networks and Modulation Spectrum Dynamics

Santos JF, Falk TH. Blind Room Acoustics Characterization Using Recurrent Neural Networks and Modulation Spectrum Dynamics. In: AES Conference: 60th International Conference: Dereverberation and Reverberation of Audio, Music, and Speech. Audio Engineering Society; 2016.pp. 3-1:1-8. Available from: https://aes.org/publications/elibrary-page/?id=18074

@inproceedings{Santos2016_18074,
  author = {Santos, João F. and Falk, Tiago H.},
  title = {{Blind Room Acoustics Characterization Using Recurrent Neural Networks and Modulation Spectrum Dynamics}},
  booktitle = {AES Conference: 60th International Conference: Dereverberation and Reverberation of Audio, Music, and Speech},
  pages = {3-1:1-8},
  year = {2016},
  month = jan,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=18074}
}

TY  - CPAPER
TI  - Blind Room Acoustics Characterization Using Recurrent Neural Networks and Modulation Spectrum Dynamics
AU  - Santos, João F.
AU  - Falk, Tiago H.
T2  - AES Conference: 60th International Conference: Dereverberation and Reverberation of Audio, Music, and Speech
SP  - 1
EP  - 8
M1  - Paper 3-1
PY  - 2016
DA  - 2016/01/06
UR  - https://aes.org/publications/elibrary-page/?id=18074
PB  - Audio Engineering Society
LA  - en
AB  - Reverberation affects perceived quality and intelligibility of speech signals, as well as the performance of automatic speech recognition systems. Having access to room acoustics characteristics of an environment may be used to improve speech processing systems, but such information is rarely available and in most cases has to be estimated blindly. Techniques based on the effects of reverberation on the modulation spectrum have been explored in the past, but they rely on its long-term average and do not use any information related to its temporal dynamics. In this paper, we aim to extract this information from the modulation spectrum time-series by using a deep recurrent neural network. We show the proposed model outperforms state-of-the-art benchmark models as well as other test models using the same signal representation in the majority of examined conditions, even when moderate amounts of noise are added to the reverberant signals.
ER  -