Speech in human system interaction

Abstract

For modern human system interaction, traditional methods based on keyboard and mouse use do not prove sufficient. Especially when thinking about the poor and ill-qualified citizens of the Information Society, we must try to find the easiest and most comfortable method for human system interaction. The optimal method for communication processes going from man to machine is (or should be) speech communication. For most people, speech is the most natural and most comfortable method of communication. In the case of control systems, speech communication offers free-hand manipulation, which sometimes can be very important. This method of human system interaction is particularly convenient for disabled persons or for ill patients as it almost guarantees their full independence and self-service (when using special robot systems for the whole day maintenance). Moreover, the proposed speech control is also fully effective in dark or when the user is in an uncomfortable position and it can also undergo (in special cases) mechanical shocks, vibrations and G-force. Last but not least, we must underline as well that such a form of communication is for almost every human being easy to learn and very natural in use. Using the speech signal as an information carrier between man and machine we can also use all devices designed for speech communication between people (e.g. phones, both stationary and mobile) which can be used as the cheapest and easiest form of remote control. Moreover, in case of an emergency, the vocal reaction of almost every human being can be much faster and more precise than any other forms of communication and steering manipulations. In this paper, some general remarks about the speech human system interaction are presented and some problems connected with this area of technological activity are discussed. On the basis of such discussion the survey of speech recognition systems is shown and a compact analysis of the most important and current scientific and technological problems related to such systems is discussed. Final conclusions are directed both towards the presentation of existing speech recognition systems and towards the forecast of solutions, available in the near future.