The term virtualizer is often used to refer to audio systems that convert multi-channel audio inputs to 2 ch. audio inputs by artificially creating a set of virtual speakers at selected positions around the listener to render a 5.1 ch. (or 7.1 ch.) home theatre system out of just 2 speakers in the front. A virtual speaker is usually generated via convolution with the head-related transfer function associated with the position of each surround speaker and appropriate filtering to eliminate cross-talk. Given a virtualizer incorporated in a TV or PC, however, it is difficult to assess if the system effectively places the virtual speakers at desired positions and thereby conclude which algorithm is “better” in terms of positioning capability. A subjective listening test involving many individuals is time consuming and costly. Therefore, it is necessary to devise a quantitative evaluation technique involving a dummy head microphone system rather than a group of individuals to provide a quick measure of speaker positions to audio engineers who develop virtualizer algorithms. In this paper, a method to track the azimuth of a virtual speaker from the output signals of a B&K HATS is proposed and the evaluation results on 3 selected virtualizer algorithms are shown.