Speech Technology: Past, Present and Future

Speech technology is intended to enable human-machine communication by voice. This includes making computers able to understand human speech, speech recognition, to produce intelligible and natural sounding speech, speech synthesis, and to determine who is speaking, speaker recognition. Unlike many other technology areas, speech technology is closely linked to an understanding of basic human properties. Without knowledge of the fundamentals of human speech production and perception, progress in speech technology will be difficult to achieve. In this paper we start out with an overview of human speech communication before describing the main trends in the development of speech technology, from the early efforts of mimicking human speech production up till today�s, mainly data-driven statistical approaches. Although the progress in speech technology performance has been great over the last 25 years, the technology is far from perfect. Many useful products and services employing speech technology exist today, but significant research issues need to be resolved before computers can approach the performance of human listeners and talkers.