Directivity steering principle for biomimicry silicon microphone

In this paper, we propose a novel sound source separation principle particularly suitable for the biomimicry silicon microphone. Conventionally, sound source separation has been performed by a wavelength-sized array of pressure-sensitive microphones. But it requires a long observation interval and much computational cost due to the learning scheme of the algorithm, necessity for statistical properties of sound, subband decomposition or FFT, and so on. While, in our biomimicry microphone, a tiny gimbal supported diaphragm detects sound pressure and its spatial gradients simultaneously. We show that their weighted summation generates sensitivity zones invariant to the sound frequency, which enables us to exclude noise sources with high temporal resolution without distorting the target sound. We theoretically describe the principle and confirm the efficiency by some experiments.