This paper presents the theoretical background to various strategies for inverse filtering of room acoustic response. Results are also presented of objective and subjective experiments designed to evaluate the effectiveness of a signal processing system which produces virtual acoustic sources in automotive interiors. The emphasis in this work was on finding the scheme which creates the most robust solution with respect to localisation accuracy, as well as the size of the 'sweet-spot', in addition to the clarity and the minimal colouration of the reproduced signals. It is shown that by inverting only the first wavefront and reflections of the impulse responses in the car, captured in a free-field model (which includes the inner parts of the doors, the driver seat and the dummy-head) in an anechoic chamber, better performance is achieved compared with the other inversion strategies investigated.