A New Bayesian Method Incorporating With Local Correlation for IBM Estimation

	A New Bayesian Method Incorporating With Local Correlation for IBM Estimation
	Liang, Shan; Liu, Wenju; Jiang, Wei
发表期刊	IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
	2013-03-01
卷号	21 期号:3 页码:476-487
文章类型	Article
摘要	A lot of efforts have been made in the Ideal Binary Mask (IBM) estimation via statistical learning methods. The Bayesian method is a common one. However, one drawback is that the mask is estimated for each time-frequency (T-F) unit independently. The correlation between units has not been fully taken into account. In this paper, we attempt to consider the local correlation information from two aspects to improve the performance. On one hand, a T-F segmentation based potential function is proposed to depict the local correlation between the mask labels of adjacent units directly. It is derived from a demonstrated assumption that units which belong to one segment are mainly dominated by one source. On the other hand, a local noise level tracking stage is incorporated. The local level is obtained by averaging among several adjacent units and can be considered as an approach to true noise energy. It is used as the intermediary auxiliary variable to indicate the correlation. While some secondary factors are omitted, the high dimensional posterior distribution is simulated by a Markov Chain Monte Carlo (MCMC) method. In iterations, the correlation is fully considered to compute the acceptance ratio. The estimate of IBM is obtained by the expectation. Our system is evaluated and compared with previous Bayesian system, and it yields substantially better performance in terms of HIT-FA rates and SNR gain.
关键词	Bayesian Rule Computational Auditory Scene Analysis (Casa) Ideal Binary Mask (Ibm) Markov Chain Monte Carlo (Mcmc)
WOS标题词	Science & Technology ; Technology
关键词[WOS]	VOICED SPEECH SEGREGATION ; SPECTRAL SUBTRACTION ; NOISE ; INTELLIGIBILITY ; RECOGNITION ; SEPARATION ; ALGORITHM
收录类别	SCI
语种	英语
WOS研究方向	Acoustics ; Engineering
WOS类目	Acoustics ; Engineering, Electrical & Electronic
WOS记录号	WOS:000313425100002
引用统计	被引频次：12[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://ir.ia.ac.cn/handle/173211/2976
专题	多模态人工智能系统全国重点实验室_机器人视觉
作者单位	Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
第一作者单位	模式识别国家重点实验室
推荐引用方式 GB/T 7714	Liang, Shan,Liu, Wenju,Jiang, Wei. A New Bayesian Method Incorporating With Local Correlation for IBM Estimation[J]. IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING,2013,21(3):476-487.
APA	Liang, Shan,Liu, Wenju,&Jiang, Wei.(2013).A New Bayesian Method Incorporating With Local Correlation for IBM Estimation.IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING,21(3),476-487.
MLA	Liang, Shan,et al."A New Bayesian Method Incorporating With Local Correlation for IBM Estimation".IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING 21.3(2013):476-487.

条目包含的文件		下载所有文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
LS_TSLP.pdf（2441KB）	期刊论文	作者接受稿	开放获取	CC BY-NC-SA	浏览下载