Along with the popularization of cellular phone, more and more people are using text messages to cell phone. However a method of inputting characters using buttons is difficult to do. As an alternative to that method, sound which training isn't required is preferable. A monosyllabic that free of capacity, such as a standard tem-plate, is ideal for use in voice-recognition system. As the preprocessing of syllable recognition, this paper proposes a method of dividing monosyllabic into phonemes. If a method using floating-point arithmetic is utilized, 1C for cell phone is large in size. Therefore, two proposed methods that use amplitude and frequency are composed of algorithms without using it. As a result of experiment by computer simulations, it is shown that both proposed methods in segment accuracy are better than conventional methods.
展开▼