Investigation of Speech Separation as a Front-End for Noise Robust Speech Recognition

By Arun Narayanan, Deliang Wang. 2014

136 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Tara N. Sainath
2. Bo Li
3. Rohit Prabhavalkar
4. Jinyu Li
5. Khe Chai Sim
6. Arun Narayanan
7. Dong Yu
8. Chin-Hui Lee

In Speech & Audio

1. Tara N. Sainath
2. Bo Li
3. Rohit Prabhavalkar
4. Deliang Wang
5. Arun Narayanan
6. Jinyu Li
7. Khe Chai Sim
8. Chin-Hui Lee

In Signal Processing

1. Bo Li
2. Jinyu Li
3. Rohit Prabhavalkar
4. Tara N. Sainath
5. Arun Narayanan
6. Dong Yu
7. Khe Chai Sim
8. Deliang Wang

Citations in the corpus

• Toward Domain-Invariant Speech Recognition via Large Scale Training (2018, background): “In fact, the gains over MTR with specialized feature enhancement is usually minimal [6, 5].”
• A Regression Approach to Speech Enhancement Based on Deep Neural Networks (2015, background): “See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. a greedy layer-wise unsupervised learning algorithm.”
• Automatic gain control and multi-style training for robust small-footprint keyword spotting with deep neural networks (2015, background): “, creating training instances by artificially adding in noise to simulate expected evaluation conditions, which has been shown to result in large improvements while recognizing noisy speech [9, 10, 11].”
• A Spectral Masking Approach to Noise-Robust Speech Recognition Using Deep Neural Networks (2014, methodology): “With the fast adoption of DNNs in various machine learning tasks, the original support vector machines used for mask estimation were replaced by DNNs [23], [24].”
• Joint noise adaptive training for robust automatic speech recognition (2014, methodology): “The IRM is estimated using a system similar to [8], but is simplified so that it can be easily incorporated 1[8] performs online feature adaptation, whereas [10] does not. into the joint framework.”
• Improving Robustness of Deep Neural Network Acoustic Models via Speech Separation and Joint Adaptive Training (2015, methodology): “The ideal ratio mask, which is defined as the ratio of the clean signal energy to the noisy signal energy at each T-F unit [37], is used in [29], and is estimated using DNNs.”