2006
DOI: 10.1093/ietisy/e89-d.3.922
|View full text |Cite
|
Sign up to set email alerts
|

A Non-stationary Noise Suppression Method Based on Particle Filtering and Polyak Averaging

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
11
0

Year Published

2007
2007
2017
2017

Publication Types

Select...
4
3

Relationship

0
7

Authors

Journals

citations
Cited by 13 publications
(11 citation statements)
references
References 0 publications
0
11
0
Order By: Relevance
“…The speech recognition system used in VoiceTra consisted of a frontend, which performed noise suppression using particle filtering [13] and acoustic analysis, and a backend, which performed large vocabulary continuous speech recognition (LVCSR) using ATRASR [14].…”
Section: Multilingual Speech Recognitionmentioning
confidence: 99%
“…The speech recognition system used in VoiceTra consisted of a frontend, which performed noise suppression using particle filtering [13] and acoustic analysis, and a backend, which performed large vocabulary continuous speech recognition (LVCSR) using ATRASR [14].…”
Section: Multilingual Speech Recognitionmentioning
confidence: 99%
“…For analyzing the speech data, we employ an HMM-based speech recognizer (ATRASR 11 developed by Itoh et al), which is high-precision speech recognition software on noise environments 12 to obtain the phoneme segmentation. The phoneme segmentation result is further converted to the viseme segmentation using a phoneme-viseme mapping table by the simple table lookup The supported languages are Japanese and English.…”
Section: Viseme Segmentation Servermentioning
confidence: 99%
“…Our proposed method used an extended Kalman particle filter with residual sampling and MCMC as did [6]. To introduce it to MM-NS, the distributions of noise models are used as priors for particles.…”
Section: Noise Suppression Proceduresmentioning
confidence: 99%
“…The Gaussian Mixture Model (GMM) based Minimum Mean-Squared Error (MMSE) method [4] assumes that input noise is stationary but fluctuating. Recently, noise suppression research has focused on non-stationary noise, including a sequential EM approach [5], a particle filtering approach [6], and so on. Since these methods usually assume that only one kind of noise signal exists, applying them to noisy speech that includes many kinds of noise signals is difficult.…”
Section: Introductionmentioning
confidence: 99%