Reasons why current speech-enhancement algorithms do not improve speech intelligibility and suggested solutions

By P. Loizou, Gibak Kim. IEEE Transactions on Audio, Speech, and Language Processing, 2011

313 citations recorded by the source corpus.

Existing speech enhancement algorithms can improve speech quality but not speech intelligibility, and the reasons for that are unclear. In the present paper, we present a theoretical framework that can be used to analyze potential factors that can influence the intelligibility of processed speech. More specifically, this framework focuses on the fine-grain analysis of the distortions introduced by speech enhancement algorithms. It is hypothesized that if these distortions are properly controlled, then large gains in intelligibility can be achieved. To test this hypothesis, intelligibility tests are conducted with human listeners in which we present processed speech with controlled speech distortions. The aim of these tests is to assess the perceptual effect of the various distortions that can be introduced by speech enhancement algorithms on speech intelligibility. Results with three different enhancement algorithms indicated that certain distortions are more detrimental to speech intelligibility degradation than others. When these distortions were properly controlled, however, large gains in intelligibility were obtained by human listeners, even by spectral-subtractive algorithms which are known to degrade speech quality and intelligibility.

Subfields: Cognitive Neuroscience, Computational Mechanics, Signal Processing

View Paper: https://www.semanticscholar.org/paper/528ccbfde64a31ab80a8839bc9ed4bfc798804cd

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

1. John F. Woodruff
2. Nicoleta Roman
3. Wen Wang
4. Deliang Wang

In Computational Mechanics

1. John F. Woodruff
2. Nicoleta Roman
3. Deliang Wang

In Signal Processing

1. John F. Woodruff
2. Dongchao Yang
3. Qichen Ye
4. Bowen Cao
5. Yuexian Zou
6. Nicoleta Roman
7. Wen Wang
8. Deliang Wang

Citations in the corpus

• NADiffuSE: Noise-aware Diffusion-based Model for Speech Enhancement (2023, methodology): “The discriminative methods learn non-linear mappings[4] or estimate time-frequency masks [5]–[7] through annotated speech data pairs, but suffer from non-linear artifacts and poor generalizations [8].”
• Supervised Speech Separation Based on Deep Learning: An Overview (2017, unknown): “With the IBM’s ability to elevate human speech intel- 372 ligibility and its connection to the articulation index (AI) [114] 373 – the classic model of speech perception – the HIT−FA rate 374 has been suggested as an evaluation metric with the IBM as the 375 reference [97].”
• Speech intelligibility in reverberation with ideal binary masking: effects of early reflections and signal-to-noise ratio threshold. (2013, result): “These studies have both validated the IBM as a computational goal of separation algorithms, and alongside other recent work (Loizou and Kim, 2011), have helped to illustrate why measures of algo-rithm performance based on signal-to-noise ratio (SNR) (Quackenbush et al. , 1988; Greenberg et al. ,…”