Non-Stationary Prediction for Addressing the Non-Causality Problem in Fixed-Filter ANC Headphones for Speech Reduction

In some situations, speech can be a disturbing source of ambient noise. Active noise control (ANC) systems have difficulties in dealing with speech due to its non-stationary nature when the non-causality problem arises in such systems, which requires the optimal filters to be non-causal. The non-causality problem is due to the delay incurred by, e.g., digital processing or acoustic propagation paths. We propose a new fixed-filter feedforward ANC system, HCMP-ANC, which aims at attenuating speech in, e.g., office environments. Notably, it comprises a non-stationary harmonic chirp model-based prediction of speech ahead in time, thus overcoming the aforementioned delay. The results show that HCMP-ANC can outperform conventional adaptive feedforward ANC, for delays in the order of tens of samples at a sampling frequency of 8 kHz. By accounting for speech non-stationarity, HCMP-ANC can attenuate female speech in a wider frequency range of up to 3 kHz, while the conventional ANC is limited to 1.5 kHz.