Ideal ratio mask estimation using deep neural networks for robust speech recognition

By Arun Narayanan, Deliang Wang. 2013

559 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

In Speech & Audio

In Signal Processing

Citations in the corpus

Toward Domain-Invariant Speech Recognition via Large Scale Training (2018, background): “Specialized techniques like masking [9] and beamforming [10] (in the case of multi-microphone input) help in certain conditions.”
Speech Denoising with Deep Feature Losses (2018, background): “However, most pipelines still operate in the spectrogram domain [4, 5, 6, 7, 8, 9, 10, 11].”
Exploring Speech Enhancement with Generative Adversarial Networks for Robust Speech Recognition (2017, methodology): “Besides the conventional enhancement techniques [3], deep neural networks have been widely adopted to either directly reconstruct clean speech [7, 8] or estimate masks [9, 10, 11] from the noisy signals.”
A Hybrid DSP/Deep Learning Approach to Real-Time Full-Band Speech Enhancement (2017, methodology): “Deep learning techniques are already being used for noise suppression [5], [6], [7], [8], [9].”
Extracting Domain Invariant Features by Unsupervised Learning for Robust Automatic Speech Recognition (2018, background): “To alleviate this issue, some robust ASR research focuses on mapping the out-of-domain data to in-domain data using enhancementbased methods [5, 6, 7], which generally requires parallel data from both domains.”
Speech Enhancement in Multiple-Noise Conditions Using Deep Neural Networks (2016, background): “Mel-frequency spectrum [14] [16], ideal binary mask, ideal ratio mask, short-time Fourier transform magnitude and its mask [21] [22], log-power spectra are all potential candidates.”
Joint Optimization of Masks and Deep Recurrent Neural Networks for Monaural Source Separation (2015, background): “The right of Figure 1 is an L intermediate layer DRNN with full temporal connections (called stacked RNN (sRNN) in [25]).”
Cracking the cocktail party problem by multi-beam deep attractor network (2017, methodology): “ction method described in Sec. 2.3 did not work here. We also included the single and multi-channel ideal-ratio-mask (IRM) system for comparison, where a mixture spectrogram was masked by oracle IRMs [44] for each target speaker, and converted to a time domain signal with noisy phase information. As”
Noise Robust TTS for Low Resource Speakers using Pre-trained Model and Speech Enhancement (2020, methodology): “Recently, neural network-based methods using masks have achieved excellent results in tasks such as speech enhancement and speech separation, and have become mainstream [12, 13].”
Large-Scale Domain Adaptation via Teacher-Student Learning (2017, background): “This improvement is much larger than what can be obtained from the traditional feature mapping [28][29] and mask learning [30][31] methods.”
Exploring the Best Loss Function for DNN-Based Low-latency Speech Enhancement with Temporal Convolutional Networks (2020, background): “Over the last years, time-frequency masking based on the short-time Fourier transform (STFT) has been widely used for DNN-based speech enhancement [2–6].”
Separate Anything You Describe (2023, unknown): “While direct mapping on the audio spectrogram for audio source separation has demonstrated success [29], the mask-estimation-based methods, such as the estimation of ideal ratio masks [30], is still the most widely used approach in the frequency domain source separation.”
Improving Robustness of Deep Neural Network Acoustic Models via Speech Separation and Joint Adaptive Training (2015, methodology): “Apart from mask estimation, binary [42] or ratio [27], [29], [13], they have also been used in other ways.”
Multi-objective learning and mask-based post-processing for deep neural network based speech enhancement (2017, methodology): “However, other DNN-based methods, such as binary or soft mask [24, 25] based speech enhancement, assume that speech and noise are independent [12] at each time-frequency (T-F) unit.”
Sergan: Speech Enhancement Using Relativistic Generative Adversarial Networks with Gradient Penalty (2019, background): “However, these masks only modify the magnitude spectra of the input signal and ignore the phase mismatch between the noisy and clean speech signals [1,7].”
MEAD: A Large-Scale Audio-Visual Dataset for Emotional Talking-Face Generation (2020, methodology): “For audio speech content, we follow the phonetically-diverse speech corpus TIMIT [16], which is originally designed in automatic speech recognition [19,40,50].”