A zoom drive noise reduction technique for digital cameras

Abstract

Sound recorded in compact digital cameras includes a zoom drive noise caused by motor driving, gear driving, and so on. The spectral subtraction (SS) method is suitable for removing these noises because it does not require any extra hardware, but the original signal is quite distorted. Therefore we have developed an advanced SS (A-SS) method suppressing the distortion of the original signal. It first subtracts the assumed noise spectrum from an input signal, and then processes the spectrum so as to keep the power constant. In addition, digital cameras have a unique problem that the ratio of noise against a recorded signal varies frequently, which is caused by the lens variability and recording environment difference. To counter this issue, we have developed the two algorithms: the statistically optimized algorithm to pre-compute the noise spectrum beforehand, and the feedback algorithm to automatically control the quantity of subtraction according to the noise ratio. We implement the A-SS method applying the proposed algorithms into a prototype camera and verified the effectiveness on the noise reduction.