Speech automatic segmentation algorithm of audio publication with adaptive threshold adjustment

Zhang, Junxing; Shi, Lixin; Wang, Dusheng

doi:10.3724/sp.j.1087.2010.00567

Search citation statements

Order By: Relevance

Paper Sections

Select...

Requirements Of Hci4cs1

Citation Types

Supporting

Mentioning

Contrasting

Year Published

2016

Publication Types

Select...

Other1

Relationship

Self Cite0

Independent1

Authors

Journals

Cited by 1 publication

(1 citation statement)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The common methods of speech endpoint detection are time domain, frequency domain and model matching [5][6][7]. In the process of recording, however, there are uncontrollable factors, such as breathing, cough, turning page or other noises, so noise in the speech signal must be handled.…”

Section: Requirements Of Hci4csmentioning

confidence: 99%

A Tool for Cutting Large Speech Corpora: HCI4CS

Guo¹,

Bai²,

Su³

et al. 2016

Proceedings of the 2016 International Conference on Artificial Intelligence: Technologies and Applications

View full text Add to dashboard Cite

Abstract-There are two methods to cut large speech corpora, include traditional manual segmentation and machine automatic segmentation. The quality of segmentation can be controlled easily using traditional manual segmentation. However, the shortcomings of manual segmentation were also obviously such as inefficiency, high cost. As we all know, the method of machine automatic segmentation has the advantage of high efficiency, but the fussy work to find cutting error can't be omitted. Thus, this paper developed a tool of human-computer interaction for cutting speech corpora (HCI4CS), which provides segment algorithm, parameter to control, modifying the error of automatic segmentation results and generates labeling files for HTK toolkit. The research object was one thousand speeches of Primi. Using HCI4CS, a person with low cognitive competence about cutting speech corpora can achieve nearly one hundred percent accuracy.

show abstract

Section: Requirements Of Hci4csmentioning

confidence: 99%