This paper describes noise reduction based on a bilateral filter improved for an acoustical signal. A microphone array system has been utilized in various scenes. However, when we utilize the microphone array system in life space, the problem of back ground noise arises. Although the back ground noise is not always stationary, statistically independent or disjoint, the amplitude of the noise is usually small compared to the desired signal. Based on the aspects, we pay attention to a nonlinear filter called bilateral filter and apply it to an acoustical signal. Bilateral filter is known as a nonlinear filter for the image. It smoothes images while preserving edges. We improve bilateral filter for an acoustical signal and evaluate the performance of bilateral sound denoising. The experimental results show that the proposed method can reduce not only the stationary noise but also the nonstationary dependent noise.