Analysis of Prediction Intervals for Non-Intrusive Estimation of Speech Clarity Index

Peso Parada P, Sharma D, Naylor PA, van Waterschoot T. Analysis of Prediction Intervals for Non-Intrusive Estimation of Speech Clarity Index. In: AES Conference: 60th International Conference: Dereverberation and Reverberation of Audio, Music, and Speech. Audio Engineering Society; 2016. Paper 5-2. Available from: https://aes.org/publications/elibrary-page/?id=18077

@inproceedings{PesoParada2016_18077,
  author = {Peso Parada, Pablo and Sharma, Dushyant and Naylor, Patrick A. and van Waterschoot, Toon},
  title = {{Analysis of Prediction Intervals for Non-Intrusive Estimation of Speech Clarity Index}},
  booktitle = {AES Conference: 60th International Conference: Dereverberation and Reverberation of Audio, Music, and Speech},
  note = {Paper 5-2},
  year = {2016},
  month = jan,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=18077}
}

TY  - CPAPER
TI  - Analysis of Prediction Intervals for Non-Intrusive Estimation of Speech Clarity Index
AU  - Peso Parada, Pablo
AU  - Sharma, Dushyant
AU  - Naylor, Patrick A.
AU  - van Waterschoot, Toon
T2  - AES Conference: 60th International Conference: Dereverberation and Reverberation of Audio, Music, and Speech
M1  - Paper 5-2
PY  - 2016
DA  - 2016/01/06
UR  - https://aes.org/publications/elibrary-page/?id=18077
PB  - Audio Engineering Society
LA  - en
AB  - We present an analysis of prediction intervals for a non-intrusive method to estimate the clarity index (C50). The method employed to estimate C50 is a data driven approach that extracts multiple features from a reverberant speech signal which are then used to train a bidirectional long-short term memory model which maps the feature space into the target C50 value. The prediction intervals are derived from the standard deviation of the per-frame C50 estimates. This approach was shown to provide a coverage probability of 80%, i.e. 80% of times the ground truth lies between the estimated intervals, where the interval bounds are computed by using 5.6 times the standard deviation of the per-frame estimates. This accuracy is shown to be consistent with other noisy reverberant environments.
ER  -