| 摘要 |
<p>A real-time continuous analytic speech recognition method in which decisions are taken concerning the phonetic features of phonemes which are sensed in the speech signal and classified, by comparing these phonetic features with a phonetic feature-based description of words stored in a dictionary, wherein said comparison is performed according to rules which correspond to the possible overlapping pronunciation of phonemes, the sounding of final consonants before initial vowel sounds in consecutive words, the features of background noises and transmission distortion.</p> |
