New insights into loudspeaker nonlinearities: Application to acoustic echo cancellation

Gudupudi, Leela Krishna

This thesis focuses on the analysis, identification and characterisation of nonlinear distortion in loudspeakers and its application to Nonlinear Acoustic Echo Cancellation (NAEC). The first part of the thesis aims at the derivation of the more accurate and empirical loudspeaker model that emulates the loudspeaker response for the purpose of predicting and preventing the nonlinear distortion. The work suggests that the generalized polynomial Hammerstein model (GPHM) approximates more reliable practical nonlinear loudspeaker behaviour. 

In the next part, after discussing the state-of-the-art NAEC solutions, we have presented a comprehensive performance and stability analysis of the widely used NAEC algorithms. The results demonstrated that the popular NAEC solutions perform better only in a few idealistic environments and are less competent in most of the practical acoustic environments. We then proposed a novel approach to NAEC based on Empirical Mode Decomposition (EMD), a recently developed technique for nonlinear and nonstationary signal analysis. Comparative experiments with a competitive baseline approach show that the new EMD approach achieves greater nonlinear echo reduction and faster convergence.

The next part of the work is our first step to align the analysis of nonlinear distortion in loudspeakers to its physical origins. We consider the application of Hilbert-Huang Transform (HHT) to the analysis of nonlinear distortion in loudspeakers. On the basis of the results of this work, we reported an alternative interpretation of loudspeaker nonlinearities through the cumulative effects of harmonic content and intra-wave amplitude-and-frequency modulation. These new findings may stimulate and reshape the future direction of NAEC research.

Sécurité numérique
Eurecom Ref:
© TELECOM ParisTech. Personal use of this material is permitted. The definitive version of this paper was published in Thesis and is available at :
See also: