Two-stage adaptive filtering techniques for noise cancellation in hearing aids

The authors propose two-stage adaptive structures to improve speech intelligibility in noise. The first stage removes the interference and produces a distorted version of the signal. The second stage then removes this distortion and the clean signal is obtained. The authors suggest two algorithms: an unconstrained and a constrained one for the first stage. For high signal-to-noise ratios (SNRs), both perform very well and more than 20-dB noise reduction may be achieved. For low SNR, the constrained algorithm performs better with a slight increase in the computational load