A Novel Method for Microphone Independent Speech Activity Detection

Punnoose A. K.

Pre-processing stages of a speech based information access system are often adversely impacted by the end user microphone quality. This may result in the rejection of a speech recording as a silence recording and vice versa. Parameters of the pre-processing stages need to be calibrated, independent of the end user microphone parameters. This paper discusses an approach for detecting whether a recording contains speech or not, independent of the microphone used to record the file. A pitch detection algorithm is used as the baseline and two new recording level pitch based features, pitch chunk partition ratio and average pitch chunk dynamic range are proposed. Experimental results are shown which proves the speech activity detection ability of the proposed features.

Select Volume / Issues:

• Volume 10,Issue 2,Mar.,2019

Year:

2019

Type of Publication:

Article

Keywords:

Microphone Sensitivity; Speech Vs Silence Detection; Pitch Based Features

Journal:

IJECCE

Volume:

10

Number:

2

Pages:

53-59

Month:

March

ISSN:

2249-071X