2016
DOI: 10.1109/lsp.2015.2495102
|View full text |Cite
|
Sign up to set email alerts
|

Speech Enhancement with Nonstationary Acoustic Noise Detection in Time Domain

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
32
0
9

Year Published

2016
2016
2021
2021

Publication Types

Select...
6
2

Relationship

1
7

Authors

Journals

citations
Cited by 35 publications
(41 citation statements)
references
References 19 publications
0
32
0
9
Order By: Relevance
“…Combining this observation with (26), we see that it is the 1-norm integral |μ|> Δ μ/2 |F (μ)| dμ which dominates the upper bound of the reconstruction error given by (26). This is also consistent with the theorem proved in [50] which states that |f (u) −f Δ μ ,Δ u (u)| ≤ 2 |μ|> Δ μ/2 |F (μ)| dμ is valid for all u ∈ R. There are counter-examples in the literature [7] showing that ∞ −∞ |F (μ) −F Δ μ (μ)| 2 dμ cannot be bounded above by a universal constant C times the 2-norm integral |μ|> Δ μ/2 |F (μ)| 2 dμ.…”
Section: Error Analysis For Reconstruction Without Pre-filteringmentioning
confidence: 63%
“…Combining this observation with (26), we see that it is the 1-norm integral |μ|> Δ μ/2 |F (μ)| dμ which dominates the upper bound of the reconstruction error given by (26). This is also consistent with the theorem proved in [50] which states that |f (u) −f Δ μ ,Δ u (u)| ≤ 2 |μ|> Δ μ/2 |F (μ)| dμ is valid for all u ∈ R. There are counter-examples in the literature [7] showing that ∞ −∞ |F (μ) −F Δ μ (μ)| 2 dμ cannot be bounded above by a universal constant C times the 2-norm integral |μ|> Δ μ/2 |F (μ)| 2 dμ.…”
Section: Error Analysis For Reconstruction Without Pre-filteringmentioning
confidence: 63%
“…O modelo treinado em múltiplas condições (Λ L )é então definido [4] pela coleção de todos os parâmetros estimados em (11)…”
Section: Treinamento Em Múltiplas Condiçõesunclassified
“…Estas componentes são então subtraídas ou compensadas do espectro do sinal de voz. Nosúltimos anos, técnicas de realce no domínio do tempo [8], [9], [10], [11] têm sido apresentadas como alternativa aos métodos espectrais. Nas abordagens temporais, evita-se a estimação explícita das componentes espectrais dos ruídos acústicos.…”
Section: Introductionunclassified
See 1 more Smart Citation
“…In the binary masking approach, some speech spectrum bins are retained while other ones are discarded. The binary mask can be refined by combining its computation with a speech signal estimator [9], [12]. Nevertheless, such a technique based on hard binary masking may generate musical noise, which degrades speech quality.…”
Section: Introductionmentioning
confidence: 99%