Channel-prediction based noise reduction algorithm for dual microphone smartphones

In this paper, we propose a dual-microphone noise reduction algorithm for smartphone. The algorithm estimates the coherent noise power spectral density (PSD) using an adaptive filter whose weights are updated based on the power level ratio (PLR) between the microphone signals. The PSDs of residual coherent noise and the uncorrelated diffuse noise are independently estimated and combined using the PLR. Simulation results confirm improved noise reduction performance and lower speech distortion than the conventional algorithm.