In this paper, a new Human-Computer-Interaction (HCI) method for a quadriplegic, which is controlled by clenching teeth, is proposed. By simple combination of two clenching patterns, seven instructions including rest, up, down, left and right as well as click and double click actions are made for the control of a pointing device. The control source is EMGs (electromyograms), which are generated by clenching teeth and acquired on two temporal muscles in one’s forehead. For easy-to-wear, the prototype device is designed for attaching EMG electrodes on a forehead by using a headband. Stochastic values such as difference absolute mean value are used for feature extractions and Fuzzy Min-Max Neural Network (FMMNN) is used for classifying clenching patterns. The usefulness of the proposed system is confirmed by the user test.