The main advantage of subband adaptive filters (SAF) is that they provide better steady-state
convergence for LMS type adaptive algorithms over the fullband implementations with reduced
computational complexity. SAF as are commonly applied to acoustic echo cancellation (AEC) solutions,
as discussed in Sub-Band Acoustic Echo Cancellation , as the impulse response of the system to be identified are long and have strong coloration. Sub-band adaptive filters can also be applied to noise cancellation solutions for the same reason. A noise cancellation system uses two or more microphones to sample the acoustic environment. One of the microphones is the noisy microphone signal with corrupted desired speech, and another microphone is the noise
reference microphone, designed to capture the noise source rather than the desired speech signal.

SAF are important acoustic system identification problems because the eigenvalue spread and spectral
dynamic range is relatively large due to the frequency dependent characteristics of the propagation and
absorption characteristics sound. Noise cancellation is typically applied with wideband and fullband
signals further increasing spread. The of sub-band analysis essentially splits the system into many
individual bands resulting in whitening system and dividing up the eigenvalue spread.

The top left image shows the noisy input signal, the top right shows the time domain noise cancellation
result, and bottom image shows the subband noise cancellation result. As it can be observed the
subband implementation achieves deeper convergence.

VOCAL’s noise reduction solutions are part of our comprehensive software library for secure, real-time
voice, video, fax and data communications solutions over mobile, radio, internet or any other
communication network. Contact us to discuss your noise reductions requirements with our engineering