论文部分内容阅读
旨在探索音段和作为超音段信息的声调在普通话和粤语的语音可知性上的相对价值。普通话的声调总量较少(4个),而粤语更为丰富(至少7个)。通常,相对于音段而言超音段具有冗余性,因此在语音识别中它的作用不是至关重要的。但是,在低质量的语音输入中超音段(重音,声调)相对的重要性将会增加,因为相对于音段,超音段对于噪音环境和失真处理后的语音的鲁棒性更强。同时,我们对两项假设进行检测:(1)粤语中词的声调的重要性大于普通话,因为前者有更丰富的声调系统;(2)当音段信息质量渐差时,词声调重要性相对加强。因为言语中音段和声调是相互作用的,如果音段削弱,那么人们识别语音的根据更多的会转到总量占优的声调上来。
Aims to explore the relative value of the pitch and the pitch of the information as the supersonic information in the phonetic awareness of Mandarin and Cantonese. The total number of Mandarin tones is less (4), while Cantonese is more abundant (at least 7). In general, supersonic segments have redundancy over the speech segment, so its role in speech recognition is not crucial. However, the relative importance of the supersonic (accents, tones) in low-quality speech input will increase as superstrings are more robust to noise and distorted speech than they are to speech. At the same time, we test two hypotheses: (1) the tone of Cantonese words is more important than that of Putonghua because the former has a richer system of tones; (2) strengthen. Because speech and tone are interplayed in speech, if the speech is weakened, then the basis for recognizing speech will shift to the more dominant tone.