Speech Recognition and Production by Machines

Abstract

Speech synthesis and recognition by machines has been a fascinating topic for scientists and engineers for many years. Inspired by speaking and listening machines in science fiction, we have witnessed fast and steady technology advances in recent years. In this section, we first describe linear prediction speech analysis which is the most widely adopted model in digital speech communication and speech synthesis. Concatenative speech synthesis is then discussed for text-to-speech synthesis. It is followed by a discussion on hidden Markov modeling of fundamental speech units which is the most successful technique for both speech recognition and speech synthesis. Finally we conclude our discussion by contrasting speech recognition by humans and machines.