Proposing a robust RLS based subband adaptive filtering for audio noise cancellation

The elimination or reduction of audio signal noise and interference is a significant challenge in signal processing. Researchers have introduced various methods, including those based on adaptive filters, to address this issue. However, these existing methods still exhibit limitations in effectively removing specific types of noise. In this study, we aim to develop a robust enhancement of the subband version of the adaptive filter using an adaptive framework founded on the recursive least squares (RLS) filter. Our proposed method incorporates from natural logarithm and a hyperbolic cosine loss function within the cost function, resulting in superior noise reduction compared to existing techniques for speech enhancement. Furthermore, the mean and mean-square convergence of the proposed method are demonstrated theoretically. To evaluate the performance of our approach, we conducted experiments comparing it against several other methods proposed in this field. The results obtained convincingly support the efficacy of our proposed method, highlighting its potential for advanced noise cancellation in audio signal processing.