Angular region-wise speech enhancement for hands-free speakerphone

A novel speech enhancement method using a microphone array is implemented in a hands-free speakerphone prototype. The method can emphasise sound arriving from the angular region where speakers are located (speaker region) while effectively suppressing noise located in the other angular region (noise region). The method estimates the power spectral densities (PSDs) of sound sources located in each region, then calculates a postfilter that effectively reduces the noise. In the PSD estimation, the calculation complexity is reduced to make the method suitable for real-time implementation. Experiments were conducted to examine the performance of the proposed method. A drastic dropoff occurred in the resultant directivity between the regions, while constant sensitivity was maintained within the same region.