MVA Processing of Speech Features

By Chia-Ping Chen, J. Bilmes. 2007

236 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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. Jinyu Li
2. Deliang Wang
3. Arun Narayanan
4. Yifan Gong
5. Dong Yu
6. Li Deng
7. Yuxuan Wang
8. Chin-Hui Lee

In Signal Processing

1. Jinyu Li
2. Deliang Wang
3. Arun Narayanan
4. Yifan Gong
5. Dong Yu
6. Yuxuan Wang
7. Donald S. Williamson
8. Zhuo Chen

Citations in the corpus

• Time-Frequency Masking in the Complex Domain for Speech Dereverberation and Denoising (2017, positive): “After normalization, auto-regressive moving average (ARMA) filtering is performed on the input features [5].”
• Investigation of Speech Separation as a Front-End for Noise Robust Speech Recognition (2014, methodology): “Some focus on feature extraction using robust features like RASTA [19] and AFE [13], or feature normalization [3].”
• Exemplar-Based Sparse Representations for Noise Robust Automatic Speech Recognition (2011, unknown): “On AURORA-2, multi-condition recognizers employing mean and variance normalization are known to achieve state-of-the-art performance, often outperforming much more sophisticated noise robustness techniques [32].”
• Improving Robustness of Deep Neural Network Acoustic Models via Speech Separation and Joint Adaptive Training (2015, methodology): “(2), and is smoothed using a 9th order ARMA fi lter [4].”
• Speech dereverberation and denoising using complex ratio masks (2017, methodology): “After normalization, auto-regressive moving average (ARMA) filtering is performed [20].”
• Joint noise adaptive training for robust automatic speech recognition (2014, methodology): “1 after smoothing it using a 9th order ARMA filter [18].”
• Complex Ratio Masking for Monaural Speech Separation (2015, methodology): “It has been shown in [23] that applying auto-regressive moving average (ARMA) filtering to input features improves automatic speech recognition performance , since ARMA filtering smooths each feature dimension across time to reduce the interference from the background noise.”
• On Training Targets for Supervised Speech Separation (2014, methodology): “We also employ an auto-regressive moving average (ARMA) fi lter [3], [4] to smooth temporal trajectories of all features: (1)”
• Supervised Speech Separation Based on Deep Learning: An Overview (2017, unknown): “The features are post-processed with the auto-regressive moving average (ARMA) filter [19] and evaluated with a fixed MLP based IBM mask estimator.”
• Hierarchical learning : theory with applications in speech and vision (2009, background): “A large fraction of the noise-robust speech recognition literature has traditionally centered on either front-end post-processing of standard features [73, 22, 32] or modifying recognizer model parameters to compensate for train/test noise mismatch [33], rather than the design of inherently noise-ro”
• An Overview of Noise-Robust Automatic Speech Recognition (2014, methodology): “An analysis showing why MVA works well is also presented in [86].”