This paper looks at a methodology of quantifying the speaking voice, by which temporal and spectral features of the voice are extracted and processed to create a numeric code that identifies speakers, so those speakers can be searched in a database much like fingerprints. The parameters studied include: (1) average fundamental frequency (F0) of the speech signal over time, (2) standard deviation of the F0, (3) the slope and (4) sign of the FO contour, (5) the average energy, (6) the standard deviation of the energy, (7) the spectral energy contained from 50 Hz to 1,000 Hz, (8) the spectral energy from 1,000 Hz to 5,000 Hz, (9) the Alpha Ratio, (10) the average speaking rate, and (11) the total duration of the spoken sentence.

Authors: Popolo, Peter S.; Sanders, Richard W.; Titze, Ingo R.
Affiliations: National Center for Voice & Speech; University of Iowa; University of Colorado at Denver(See document for exact affiliation information.)
AES Convention: 123 (October 2007) Paper Number: 7274
Publication Date: October 1, 2007 Import into BibTeX
Subject: Audio Forensics
Permalink: https://www.aes.org/e-lib/browse.cfm?elib=14332

Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!

This paper costs $33 for non-members and is free for AES members and E-Library subscribers.

Learn more about the AES E-Library

E-Library Location: (CD 123Papers)   /123/7274.pdf

Start a discussion about this paper!