Blind separation of speech mixtures via time-frequency masking

By Özgür Yilmaz, Scott T. Rickard. IEEE Transactions on Signal Processing, 2004

1,625 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/aeab027c39a5b58c7c69addf87332c5b144205f5

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Handbook of Blind Source Separation: Independent Component Analysis and Applications (2010, background): “Some ten years elapsed before these techniques began to be fully exploited for blind source separation [72, 106, 66, 70, 116, 61, 15, 111].”
LINEAR PREDICTION-BASED DEREVERBERATION WITH ADVANCED SPEECH ENHANCEMENT AND RECOGNITION TECHNOLOGIES FOR THE REVERB CHALLENGE (2014, methodology): “DOLPHIN efficiently combines conventional direction-of-arrival (DOA) [19, 20] feature based enhancement [21] and spectral feature based approaches through a dominant source index (DSI) that indicates whether noise or speech is dominant at each time/frequency bin.”
Comparing Measures of Sparsity (2008, background): “There has also been research in the uniqueness of sparse solutions in overcomplete representations [22], [23].”
Subjective and Objective Quality Assessment of Audio Source Separation (2011, background): “Independent Component Analysis (ICA) in [13], time-frequency masking in [10] or simulated separation in [16], and a narrow range of sound material, e.”
Model-Based Expectation-Maximization Source Separation and Localization (2010, methodology): “The simplification made by DUET [4], [6] is to ignore high frequencies in which the cross-correlation is multimodal and to convert interaural phase”
Learning to Separate Object Sounds by Watching Unlabeled Video (2018, methodology): “Some methods assume access to multiple microphones, which facilitates separation [19, 51, 75].”
Under-Determined Reverberant Audio Source Separation Using a Full-Rank Spatial Covariance Model (2009, methodology): “This increases the weight of time–frequency bins of large amplitude where the modeled source is more likely to be prominent, in a way similar to [2].”
Blind Source Separation and Independent Component Analysis: A Review (2005, methodology): “In order to get more sparse signals, they employ linear transforms such as the short-time Fourier transform (STFT), and they assume that only one independent component is dominant at any given point in the timefrequency domain [148, 15, 124].”
Direction of Arrival Based Spatial Covariance Model for Blind Sound Source Separation (2014, positive): “We can approximate the SCMs being additive since the sources are approximately uncorrelated but also sparse, meaning that only a single source is to be active within each time-frequency point [33].”
Machine learning in acoustics: Theory and applications. (2019, background): “Based on the W-disjoint orthogonality property of the speech signal, 131,132 these features can be attributed a GMM (35), with each Gaussians associated with a candidate position in the enclosure on the predeﬁned grid.”
A Consolidated Perspective on Multimicrophone Speech Enhancement and Source Separation (2017, positive): “In [265], a soft mask is proposed based on a mapping between the ITD values and the relative contribution of the target source.”
Wavesplit: End-to-End Speech Separation by Speaker Clustering (2020, neutral): “This classical signal processing task (Roweis, 2001; Yilmaz & Rickard, 2004; Vincent et al., 2018) has witnessed fast progress in the recent years thanks to supervised neural networks (Wang & Chen, 2018).”
Blind One-microphone Speech Separation: A Spectral Learning Approach (2004, methodology): “Third, our framework is based on segmentation of the spectrogram and, as such, distortions are inevitable since this is a “lossy” formulation [6, 4].”
Joint DOA and multi-pitch estimation based on subspace techniques (2012, background): “This assumption may be appropriate when sources consist of mixture of several speech signals having different pitches [9].”
Recognition of overlapping speech using digital MEMS microphone arrays (2013, methodology): “Since the optimum window length for timefrequency masking is about 1024–2048 samples (at a sampling rate of F = 16kHz) [14], we use an FFT of length L = 22, with a window shift of L/32.”
Speaker-Aware Neural Network Based Beamformer for Speaker Extraction in Speech Mixtures (2017, methodology): “This problem has been investigated using both single and multi-channel source separation techniques, including NMF [1], ICA [2] or clustering of spatial cues [3, 4, 5].”
Joint Optimization of Masks and Deep Recurrent Neural Networks for Monaural Source Separation (2015, neutral): “Moreover, we find it useful to further smooth the source separation results with a time-frequency masking technique, for example, binary time-frequency masking or soft time-frequency masking [1], [11], [18], [32].”
Underdetermined Convolutive Blind Source Separation via Frequency Bin-Wise Clustering and Permutation Alignment (2011, methodology): “For the design of masks , we rely on the sparseness property of source signals [17].”