2007
DOI: 10.1109/tasl.2007.894516
|View full text |Cite
|
Sign up to set email alerts
|

Noisy Speech Enhancement Using Harmonic-Noise Model and Codebook-Based Post-Processing

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2
1

Citation Types

0
29
0

Year Published

2010
2010
2020
2020

Publication Types

Select...
5
2
2

Relationship

0
9

Authors

Journals

citations
Cited by 42 publications
(29 citation statements)
references
References 27 publications
0
29
0
Order By: Relevance
“…It is noticed that Authors done the research By ignoring the phase information Wiener-like filtering [1]- [2] focus on the magnitude or a power TF, which works good but output signals with some noise as musical noise. It has been seen that frequency resolution was limited by that of the TF transform.…”
Section: Review Of Literaturementioning
confidence: 99%
See 1 more Smart Citation
“…It is noticed that Authors done the research By ignoring the phase information Wiener-like filtering [1]- [2] focus on the magnitude or a power TF, which works good but output signals with some noise as musical noise. It has been seen that frequency resolution was limited by that of the TF transform.…”
Section: Review Of Literaturementioning
confidence: 99%
“…Esfandiar Zavarehei and et;al , [1] researchers addressed the problem of lost to noise or suppressed by noise reduction and tried to enhance the performance of conventional speech by using the harmonic noise model. They worked with amplitude, frequency, and harmonicity of sub bands of speech spectrum.…”
Section: Review Of Literaturementioning
confidence: 99%
“…1͒. Harmonics regeneration can be implemented using adaptive comb filtering ͑Nehorai and Porat, 1986͒ techniques, nonlinear functions ͑Plapous et al, 2006͒, and codebook-based techniques that capitalize on the fact that the harmonic amplitudes are highly correlated ͑Chu, 2004; Zavarehei et al, 2007͒. Once the harmonic amplitudes are estimated, it is straightforward to estimate the partials adjacent to the main harmonics using the Gaussian model shown in Fig. 3.…”
Section: A Practical Implementationmentioning
confidence: 99%
“…Instead of the time-domain signal, often low-dimensional parameter vectors are used as an intermediate representation of the sources as this leads to better and faster training of the codebooks and faster separation and enhancement algorithms. Some examples of VQ-based enhancement and separation methods are [1][2][3][4] and [5][6][7][8][9], respectively. In finding the parameters of the intermediate representation, standard estimation algorithms such as the maximum likelihood estimator based on well-known metrics can be used.…”
Section: Introductionmentioning
confidence: 99%