The man-machine interaction through speech allows, in many applications, an attractive, comfortable and effective communication. However, processing requirements still make it difficult the implementation of man-machine vocal communication systems in devices with low computational power, such as mobile phone, palmtops and home equipments. This paper describes the development of a signal preprocessing library for speech recognition systems applied to devices with low computational power. Within this context, several issues have been considered, such as limited power consumption, memory and processing capacity without reducing the efficiency of the recognition process.