A method of tracking a vocal target is disclosed. An image-capturing device is employed to capture an image including the participating targets. Next, an image tracking method is used to determine the image regions where the participating targets are respectively located. The found image regions are compared with the vocal regions of each vocal targets detected by a sound-detecting device. Thus, the positions where the vocal targets are located are precisely detected. Thus, cost of the video system can be effectively reduced and the video image resolution can be effectively promoted.