In this paper we investigate the voice control of an autonomous robot in the presence of impulsive noise. We propose an original structure of the intelligent voice control system, present experimental investigation of separate modules and outline the performance of the system by the simulation example. Our approach differs from others in twofold: the noise detection is carried out by specialized artificial neural network; and the restoration of the missing speech signal is performed by using an intelligent multirate-processing scheme. The simplicity of neural network’s employment and unnecessary a priori knowledge of noise characteristics are the main merits of this approach. The employment of nonlinear re-sampling improved the precision of speech signal restoration and consequently increased the overall recognition performance.