The fast Newton transversal filter (FNTF) family of adaptive filtering algo
rithms bridges the performance gap between the standard normalized least-me
an-square (NLMS) and fast recursive least-squares (FRLS) algorithms by allo
wing the use of linear predictors with variable orders. Recently, it has be
en shown that the FNTF is an effective scheme for fullband acoustic echo ca
ncellation (AEC) of short echo paths (e.g., mobile context). In this paper,
the merits of FNTF for subband AEC of the long echo paths typically associ
ated with the use of hands-free audio terminals in offices (e.g., audioconf
erencing) are investigated. To this end, a stabilized version of FNTF is in
corporated in an oversampled subband structure based on Weaver single-sideb
and (SSB) modulation. The performance of the resulting subband AEC system i
s evaluated in terms of convergence speed and computational complexity. The
results point to some fundamental limitations of FNTF in this application.