Disclosed are a sound source orientation control apparatus and method for video surveillance. In the present invention, a sound pickup component is capable of rotating synchronously with a camera, and thus a dynamic sound source coordinate system rotating synchronously with the camera can be formed. Accordingly, an azimuth angle obtained by resolving on the basis of the dynamic sound source coordinate system is a relative azimuth angle of a sound source with respect to a lens normal of the camera. Further, closed-loop control on the relative azimuth angle can be realized by converging the relative azimuth angle into a preset angle range, and the direction of the sound source is tracked by a lens viewing angle range of the camera through closed-loop control on the relative azimuth angle. Because loop gain generated in closed-loop control on the relative azimuth angle can suppress precision errors generated in an azimuth angle resolving algorithm, the present invention can improve the accuracy of orientation without improving the precision of the resolving algorithm, so that resources consumed for running the algorithm can be saved.