Machine Recognition of Spoken Words

摘要: Publisher Summary The mechanical recognition of speech sounds is a field in which computers are now being used. This chapter discusses the present state by machines. Speech machines must work with acoustic wave as input, and therefore perform some or all processes normally province human listener. classifies various describes their properties. classification done on an articulatory basis. In addition, examines phonetic transcription speech, it indicates economical method for design cannot be treated independently occurring events. methods used machine discussed vowels, vowel-like sounds, fricatives, plosives, binary method. Present devices operate correctly only if speaker controls his/her pronunciation within narrow limits. Contemporary at classified (1) accurate any word out small group (subdivided into (a) that deal each unit (b) break words least approximately phonemic segments), (2) Machines designed to recognize large number different words. have little tolerance changes, self-adjustment features would appear necessary. Although there present, idea how this might current self-optimizing learning may prove relevant.