Speaker anonymisation using the McAdams coefficient

Patino, Jose; Todisco, Massimiliano; Nautsch, Andreas; Evans, Nicholas
Research Report RR-20-343, 28 February 2020

This report presents an alternative to speaker anonymisation that does
not require external training data and is based on a single processing block
using basic signal processing techniques. The proposed method employs
the McAdams coefficient to apply a slight contraction/expansion to the poles
derived from linear predictive coding (LPC) coefficients of speech content
on a frame-by-frame basis, consequently leading to a transformation of the
related formants.

Type:
Report
Date:
2020-02-27
Department:
Digital Security
Eurecom Ref:
6190
Copyright:
© EURECOM. Personal use of this material is permitted. The definitive version of this paper was published in Research Report RR-20-343, 28 February 2020 and is available at :

PERMALINK : https://www.eurecom.fr/publication/6190