认知与发展心理学研究室知识库

本论文主要结合目前汉语语音方面的己有研究成果，以大规模语料库为研究材料，通过心理物理实验方法建构音节知觉的多维空间模型，对汉语语境中音节的知觉机制做了初步的探讨。通过对汉语音节知觉结构的研究，来寻找声学相关物影响听者对汉语音节知觉的心理依据;给出在汉语音节层次影响知觉结果的主要声学参数，并对知觉结果与相关声学参数的关系做初步分析;进一步探讨韵律信息对音节特征的影响。
结合目前言语工程技术的发展现状，本文的研究结果也为基于拼接的语音合成系统的样本库设计、语音识别和合成系统音节调节参数的选择等方面技术的改进提供了理论参考。
本论文的主要研究结果为:
a)通过对阴平声音节“gongl”的实验研究发现，影响阴平声音节知觉结构的声学参数可以主要归结为平均音高和时长。通过对去声音节“ju4"的实验研究，结果表明，音域(或高音点)、时长、平均音高(或低音点)构成影响去声音节知觉结构的三个主要声学参数。对阴平声和去声 音节的韵律分析表明，音节在句中位置、所在韵律词长度等韵律信息决定了音节在知觉空间的分布趋势。
b)通过听辨实验考察了前后相邻音节协同发音对中间音节知觉差异的影响作用，发现协同发音引起的音节知觉差异明显小于与超音段特征变化有关的声学参数变化引起的音节知觉差异:而协同发音引起的音节知觉差异主要取决于前音节末尾韵母部分的差异、和后音节开头辅音发音部位的差异。

Based on researches related to Mandarin speech, this thesis develops a percep-tive mufti-space model of Mandarin syllables, with a method of psychological-physical experiment. Material of the experiments comes from large speech database.With the mufti-space model we have a primary research on the perceptive principle of Mandarin syllables. The anticipant aims of the thesis are: Find the evidences of acou-stir correlates affecting perception of Mandarin syllables; Put forward the main acoustic parameters which determine perceptive outcome on the level of Mandarin syllable, and have an analysis to the relation between perceptive outcome and correla-tive acoustic parameters; Have a discussion about the effect to the syllable character-ristics, of prosodic information.
In relation to the development of speech technique, the research gives a reference to the corpus design of concatenative speech synthesis system, selection of syllable adj ustment in speech synthesis system and speech recognition system, etc.
The main results of the thesis is listed as:
a) Based on the research of syllable "gongl”with tone 1，we find: the main acoustic 一factors affecting perceptive structure of syllable with tone 1 are average FO and duration. the prosodic information such as position of syllable in sentence, length of the prosodic word containing the syllable, decide syllable distribution in perceptive space. Result of experiments about syllable " ju4" with tone 4 is: range of FO (or high FO point), duration and average FO(or low FO point) are main acoustic parame-ters deciding perceptive structure of syllable with tone 4. Prosodic analysis about syllables with both tone l and tone 4 indicates, position of syllable in sentence,length of the prosodic word containing the syllable also determine syllable distribu-tion in perceptive space.
perceptive difference of middle syllables. Results show that perceptive difference of syllables caused by co-articulation is distinctly less than perceptive difference of syllables caused by diversification of acoustic parameters which are in relation to super-segmental feature changinb; perceptive difference of syllables caused by co-articulation lies on variance of both kinds of terminal part of the front syllable and pronunciation locations of initial part of the latter syllable.