Human moving detection by frequency analysis using video signal

In a motion detection area, technology such as pattern recognition through pattern matching and/or a neural network is often used. Although the algorithm "motion vector detection" for checking the motion in MPEG has been exploited, it is only for detecting the highest correlation block of a local area between adjacent frames and does not detect a motion of the same objective portion exactly. Hence it has a little lack of reliability for the motion detection. In this research the difference between the continuation frames of a video picture is utilized. The DCT of the difference can be implemented in a very short time, the alteration of the amplitude of each frequency domain is classified as a pattern in real time, and finally the state of the motion is estimated. In this way the weak point of the conventional method is also avoidable.