We investigated on the possibility of an active cancellation system of unnecessary speech radiation. The intended application of this system is to cancel cellular speech, and also to cancel speech input for recognition-based dictation or speech-command control systems. Both of these applications do not require speech to be radiated into surrounding space, only into the input microphone, and would benefit if global radiation is controlled. Through simulations, it was found that speech cancellation is possible with a secondary source placed close to the mouth generating inversed-phase speech. The inversed-phase speech can be predicted using linear prediction. It was found that 1) over 14 dB cancellation of speech is possible, 2) the cancellation is greater when the distance between the mouth and the secondary sound source is smaller, and 3) higher order LPC yields greater cancellation up to an order of 128. We plan to implement a prototype system using DSPs.