Evaluation of Objective Quality Measures for Speech Enhancement

By Y. Hu, P. Loizou. 2008

1,990 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Computational Mechanics, 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 Cognitive Neuroscience

In Computational Mechanics

In Signal Processing

Citations in the corpus

SEGAN: Speech Enhancement Generative Adversarial Network (2017, methodology): “During train, we extract chunks of waveforms with a sliding window of approximately one second of speech (16384 samples) every 500ms (50% overlap).”
Real Time Speech Enhancement in the Waveform Domain (2020, background): “(v) COVL: MOS prediction of the overall effect [26] (from 1 to 5).”
ViSQOL: an objective speech quality model (2015, methodology): “This was shown to be the case by Hu and Loizou when evaluating speech enhancement algorithms with a variety of objective quality metrics [29].”
Interactive Speech and Noise Modeling for Speech Enhancement (2020, background): “• CBAK (Hu and Loizou 2007): MOS prediction of the in-trusiveness of background noises (from 1 to 5).”
An Overview of Deep-Learning-Based Audio-Visual Speech Enhancement and Separation (2020, background): “CSIG / CBAK / COVRL [104] 2007 Composite measures which combine [108] basic objective measures”
Speaker-Aware Neural Network Based Beamformer for Speaker Extraction in Speech Mixtures (2017, methodology): “The accuracy was evaluated on the beamforming outputs with speech enhancement measures — cepstral distance (CD) [27] and signal-to-distortion ratio (SDR) [28].”
Polyglot Neural Language Models: A Case Study in Cross-Lingual Phonetic Representation Learning (2016, background): “The MCD metric calculates an L2 norm of the Mel Frequency Cep-stral Coefﬁcients (MFCCs) of natural speech from a held out test set, and synthetic speech generated from the same test set.”
Deep Residual-Dense Lattice Network for Speech Enhancement (2020, background): “CSIG,CBAK, andCOVL are mean opinion score (MOS) predictors of the signal distortion, background-noise intrusiveness, and overall signal quality, respectively (Hu and Loizou 2008).”
Improved Speech Enhancement with the Wave-U-Net (2018, neutral): “These are: 97 CSIG, a prediction of the signal distortion attending only to the speech signal [3] (from 1 to 5); CBAK, 98 a prediction of the intrusiveness of background noise [3] (from 1 to 5); and COVL, a prediction of 99 the overall effect [3] (from 1 to 5).”
PHASEN: A Phase-and-Harmonics-Aware Speech Enhancement Network (2019, background): “• CSIG (Hu and Loizou 2007): Mean opinion score (MOS) prediction of the signal distortion attending only to the speech signal (from 1 to 5).”
An Algorithm for Intelligibility Prediction of Time–Frequency Weighted Noisy Speech (2011, result): “This result is in line with the conclusions from Hu and Loizou [26] where, in general, no noise-reduction scheme could improve the intelligibility of noisy speech.”
Complex Ratio Masking for Monaural Speech Separation (2015, background): “PESQ and SNR fw have been shown to be highly correlated to human speech quality scores [31], while STOI has high correlation with human speech intelligibility scores.”
Generative Pre-training for Speech with Flow Matching (2023, methodology): “CBAK refers to composite background intrusiveness (Hu & Loizou, 2007).”
Fréchet Audio Distance: A Metric for Evaluating Music Enhancement Algorithms (2018, background): “Human raters are typically provided with individual audio clips of speech and asked to evaluate the naturalness of the speech signal on a five point scale from 5 (very natural, no degradation) down to 1 (very unnatural, very degraded) and how intrusive the background noise is from 5 (not noticeable)”
Conditional Diffusion Probabilistic Model for Speech Enhancement (2022, methodology): “We consider perceptual evaluation of speech quality (PESQ) [29], prediction of the signal distortion (CSIG), background intrusiveness (CBAK), and overall speech quality (COVL) [30] as the evaluation metrics.”
Subjective and Objective Quality Assessment of Audio Source Separation (2011, neutral): “are applied to the overall distortion instead of individual components [24], [25].”
SpeechBrain: A General-Purpose Speech Toolkit (2021, background): “The resulting metric showed a much higher correlation with human judgments than any contributing metric [112].”
The reverb challenge: A common evaluation framework for dereverberation and recognition of reverberant speech (2013, methodology): “The CD scores are calculated as in [12].”
Time-Frequency Masking-Based Speech Enhancement Using Generative Adversarial Network (2018, background): “CBAK and CMOS measured from 1 to 5, predicts the extent of background interferences in the speech and the overall effect, respectively [25].”