Speech Recognition and Synthesis

Speech Recognition and speech synthesis are needed to provide computers with the ability to communicate with humans in a natural way. Research has been going on in the area of speech recognition for more than 50 years and still remains an unsolved problem.
Speech synthesis is relatively simpler compared to speech recognition. Speech recognition involves digital signal processing and pattern recognition techniques. Just as we use alphabets for text, speech signals have been found to have a set of around 45 alphabets called phonemes.
The key idea is to process the speech signals to extract important parameters and then use these parameters to identify phonemes. Then map these phonemes to alphabets to recognize the spoken word or sentences. There are problems related to intonation, dialect variation, emotions, etc. affecting the spoken sentences. These issues are important in speech synthesis as well.