Single-channel speech separation using sparse non-negative matrix factorization

By Mikkel N. Schmidt, Rasmus Kongsgaard Olsson. Interspeech, 2006

411 citations recorded by the source corpus.

Subfields: Computational Mechanics, Signal Processing

View Paper: https://www.semanticscholar.org/paper/90bc7d8beae294dbd1fead2e9c613665f281d4a9

How researchers see this paper

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

In Computational Mechanics

In Signal Processing

Citations in the corpus

Multitalker Speech Separation With Utterance-Level Permutation Invariant Training of Deep Recurrent Neural Networks (2017, methodology): “Another popular technique for multi-talker speech separation is Non-negative Matrix Factorization (NMF) [11]–[14].”
Deep Neural Networks for Single-Channel Multi-Talker Speech Recognition (2015, methodology): “Model based approaches use factorial GMM-HMM [3] to model the interaction between the target and competing speech signals and their temporal dynamics [4],[5],[6].”
Single channel speech music separation using nonnegative matrix factorization and spectral masks (2011, methodology): “In [1], sparse NMF was used with trained basis vectors to separate the mixture of two speech signals.”
Recursive speech separation for unknown number of speakers (2019, background): “Previously, various approaches including spectral clustering [5] computational auditory scene analysis (CASA) [6], non-negative matrix factorization (NMF) [7–10] were proposed to tackle this problem, yet showed limited success.”
CBLDNN-Based Speaker-Independent Speech Separation Via Generative Adversarial Training (2018, methodology): “Non-negative matrix factorization (NMF) based methods [3, 4, 5] decompose signal into sets of bases and weight matrices, which are used to estimate mixing factors during evaluation.”
Discretization and Re-synthesis: an alternative method to solve the Cocktail Party Problem (2021, positive): “Compared to conventional approaches like computational auditory scene analysis [8], [9] and non-negative matrix factorization [10], end-to-end models are entirely data-driven, achieving remarkable improvement in speech quality and intelligibility.”
Single-channel mixed speech recognition using deep neural networks (2014, methodology): “Some other approaches including [9] and [10] utilize the non-negative matrix factorization (NMF) for the separation and pitch-based enhancement.”
Dual-Path Transformer Network: Direct Context-Aware Modeling for End-to-End Monaural Speech Separation (2020, methodology): “Before the deep learning era, many traditional methods are introduced for this task, such as non-negative matrix factorization (NMF) [3, 4], computational auditory scene analysis (CASA) [5] and probabilistic models [6].”
Energy Disaggregation via Discriminative Sparse Coding (2010, methodology): “, [24, 23]), which forms for the basis for our disaggregation approach, is to train separate models for each individual class Xi, then use these models to separate an aggregate signal.”
Deep neural networks for single channel source separation (2013, background): “Another approach to model the training data is to train nonnegative dictionaries for the source signals [13, 14, 15].”
Paralinguistics in speech and language - State-of-the-art and the challenge (2013, methodology): “Typical acoustic LLDs in the field cover: intonation (pitch, etc.), intensity (energy, Teager functions, etc.), linear prediction cepstral coefficients (LPCC), perceptual linear prediction (PLP) parameters, cepstral coefficients (Mel frequency cepstral coefficients (MFCCs), etc.), formants (amplitud”
Deep clustering: Discriminative embeddings for segmentation and separation (2015, background): “In the audio domain, single-channel speech separation methods, for example, segment the time-frequency elements of the spectrogram into regions dominated by a target speaker, either based on classifiers [6–8], or generative models [9–11].”
Exploring Monaural Features for Classification-Based Speech Segregation (2013, background): “Algorithms such as [19], [27], [28] can work well if the statistical properties of the observations correspond well to training conditions.”
Recent progresses in deep learning based acoustic models (2017, methodology): “e been made to attack this problem. Before the deep learning era, the most popular techniques include computational auditory scene analysis (CASA) [130]–[132], non-negative matrix factorization (NMF) [133]–[135], and model based approach [136]–[138], such as factorial GMM-HMM [139]. Unfortunately th”