Real-time acoustic noise canceling technique on innovations-based inverse kepstrum and FIR RLS

This paper presents real-time noise canceling technique using inverse kepstrum system as front-end innovations-based whitening application to FIR (finite impulse response) RLS (recursive least square) filter. Based on the identification of acoustic path transfer functions between two microphones, the ratio of acoustic path transfer function is analyzed as unknown system in ANC (adaptive noise canceling), and applied as innovation-based application to Wiener filtering structures. For the application in a realistic environment, noise canceling performance will be verified in beamforming structure, where the application of front-end innovations-based inverse kepstrum with cascaded FIR RLS could provide even better performance with reduced computational complexity compared with ordinary ANC approach using IIR (infinite impulse response) RLS filter.