Table tennis has been developed widely and promoted as a popular sport in Taiwan. It is due to the space being very limited and crowded people. Only some space is enough for the organization of a table tennis club. Since the volume of the table tennis is small and the weight is very light for Table tennis, the speed and the spin vary dramatically in a real game. Thus it is attractive for people to enjoy this sport. Table tennis also has been regarded as a major sport in many countries all over the world. In addition, it has been include as a formal competition in the Olympic game. A referee stays near the table for fair judgment. However, she/he blocks the game in some viewpoints for the audience. To develop an automatic method for on-table ball detection is a good idea. In this thesis, we attempt to identify an on-table ball event by using the detection of pulse and of peak spectrum for a sound feature in the frequency domain. In order to improve the accuracy of on-table ball detection, the sound events of players’ shoes friction on floor and the noise produced by the audiences and the environment should be rejected. Experimental results show that the proposed method can efficiently detect the on-table ball event in running a real table tennis game.
