Towards Scaling Up Classification-Based Speech Separation

By Yuxuan Wang, Deliang Wang. IEEE Transactions on Audio, Speech, and Language Processing, 2013

472 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

View Paper: https://www.semanticscholar.org/paper/6f22bea906bb4fa0759da6954aaf0fdf860d4688

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 Signal Processing

Citations in the corpus

Deep neural network based spectral feature mapping for robust speech recognition (2015, background): “Deep neural networks (DNNs) have shown strong learning capacity [1] and been successfully applied to many speech applications [2, 3].”
A Regression Approach to Speech Enhancement Based on Deep Neural Networks (2015, methodology): “A simple yet effective method to cope with the unseen noise conditions is to include many different noise types in the training set [15, 27].”
A deep neural network for time-domain signal reconstruction (2015, methodology): “In this work, we propose a new deep network that directly reconstructs the time-domain clean signal through an inverse fast Fourier transform layer.”
Investigation of Speech Separation as a Front-End for Noise Robust Speech Recognition (2014, methodology): “We use DNNs 2 to estimate the IRM as they show good performance [43] and training using stochastic gradient descent scales well compared to other nonlinear discriminative classi ﬁ ers, such as SVMs, as the size of the dataset increases [2].”
Deep learning for monaural speech separation (2014, methodology): “The authors first try K neural networks to predict each feature dimension separately, where K is the feature dimension, and then train another classifier (one layer perceptron [13] or an SVM [14]) using neighboring time-frequency predictions in the first stage as the input.”
Multitalker Speech Separation With Utterance-Level Permutation Invariant Training of Deep Recurrent Neural Networks (2017, background): “More recently, a large number of techniques based on deep learning [19] have been proposed, especially for Automatic Speech Recognition (ASR) [20]–[25], and speech enhancement [26]–[34].”
An Experimental Study on Speech Enhancement Based on Deep Neural Networks (2014, methodology): “Other data-driven methods attempt to make a binary classification decision on time-frequency (T-F) units, such as estimating the ideal binary mask for monaural speech separation [13], however the acoustic context information of the T-F unit is not well utilized in a classification framework.”
An algorithm to improve speech recognition in noise for hearing-impaired listeners. (2013, methodology): “As a result, the Kim et al. algorithm likely has difficulty handling even small noise variations between training and test conditions (Han and Wang, 2012; Wang and Wang, 2013).”
On Training Targets for Supervised Speech Separation (2014, methodology): “Supervised speech separation has also been shown to generalize well given sufficient training data [34], [36].”
High-dimensional and large-scale anomaly detection using a linear one-class SVM with deep learning (2016, methodology): “, a DBN in combination with labelled SVMs such as binary or multiclass SVMs) have been used in several application domains, such as object classification [14], text classification [27], music classification [29], face expression recognition [41], and speech separation [42].”
A Convolutional Recurrent Neural Network for Real-Time Speech Enhancement (2018, background): “Inspired by the concept of time-frequency (T-F) masking in computational auditory scene analysis (CASA) [3], speech separation has been formulated as supervised learning in recent years, where a deep neural network (DNN) is employed to learn a mapping from noisy acoustic features to a T-F mask [4].”
A New Framework for CNN-Based Speech Enhancement in the Time Domain (2019, background): “popular deep learning methods are denoising autoencoders [4], feedforward neural networks [5], [6], and CNNs [7]–[9].”
Convolutional Neural Networks to Enhance Coded Speech (2018, methodology): “Deep neural networks (DNNs) are used as a classification method to estimate the ideal binary mask [23] or smoothed ideal ratio mask [24] for noise reduction.”
Supervised Speech Separation Based on Deep Learning: An Overview (2017, unknown): “Illustration of DNN for feature learning, and learned features are then used by linear SVM for IBM estimation (from [181]).”
Deep Learning for Environmentally Robust Speech Recognition (2017, unknown): “Speciﬁcally, Wang and Wang [90] ﬁrst introduced DNNs to perform IBM estimation for speech separation, which signiﬁcantly outperforms other methods without deep learning.”
Multichannel Audio Source Separation With Deep Neural Networks (2016, background): “Although the authors in [15] considered both speech and music separation, most studies focused either on speech separation [10]–[12], [14], [17]–[21] or on music separation [13], [16].”
Ideal ratio mask estimation using deep neural networks for robust speech recognition (2013, unknown): “The IBM-direct system is most similar to the one proposed in [21], which has been shown to perform well for speech separation.”
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.”
Speaker-Independent Speech Separation With Deep Attractor Network (2017, background): “of the sources given the T-F representation of the audio mixture (such as noisy speech or multiple speakers) [22], [23], [27], [28].”