Abstract
A model for speech and noise sound fields in a car cabin is proposed. The model takes into account a complicated geometry of the cabin, frequency dependence of sound absorption on its surfaces and spatial correlation of distributed noises. The sound field computation is based on the wave approach for the range of low frequencies (below 1000 Hz) and on the ray approximation for higher frequencies. The software is developed, which provides simulation of speech and noise fields in the cabin during the car motion. This software may be utilized as an simulator of the sound field for the development of speech coding and recognition devices intended for car cabins. To check the model adequacy the spectra of simulated noise of the engine and results of natural measurements are compared. The effectiveness of microphone arrays as speech enhancement tools is investigated on the base of the developed model for different signal processing algorithms and noise conditions.