您的位置:首页  > 论文页面

汉语单音节和双音节词感知清晰度的测算方法研究

发表时间:2013-11-30  浏览量:1343  下载量:505
全部作者: 姜旭栋,朱维彬
作者单位: 北京交通大学计算机与信息技术学院人机交互工程;北京交通大学信息科学研究所
摘 要: 目的:解决临床言语听力测试材料的等价性评估问题。方法:模拟人的听音感知过程,构建一个能定量出常用基本测试单元(单音节、双音节词等)感知清晰度的测算模型。模型以条件概率模型为基础,除了声母音位和韵母音位,细化考察了声调、音联、词频等因素对言语感知清晰度的影响。通过组织单音节和双音节词的测听试验,获得全通,LP1600,HP1600等传递条件下的音联音位混淆矩阵和清晰度实测值,确定感知清晰度模型的频度影响因子。结果:通过各模型计算值与试验实测值的比较可知,新提出的过程感知模型与实测值具有更小的整体偏差程度和更显著的相关性。
关 键 词: 语音处理;言语测听方法;清晰度测算;语音感知;汉语单音节与双音节
Title: Research on calculating method of perceptual articulation about Chinese monosyllable and disyllabic word
Author: JIANG Xudong, ZHU Weibin
Organization: Human-Uachine Interaction Engineering, College of Computer & Information Technology, Beijing Jiaotong University; Institute of Information Science, Beijing Jiaotong University
Abstract: Objective: To solve the problem of evaluating the equivalence of audiometric materials. Methods: The paper proposes a new speech assessment method to simulate the process of human listening perception. This paper investigates the independent contribution of initial and final based on conditional probability model, emphasizes the effect on the perceptual articulation of monosyllable, disyllabic word, which is resulted from the factors of tone, junctures and word frequency. The paper organizes a monosyllable and a disyllabic word audiometric tests to get data base just as the phonematic perceptual confusion matrix and the measured value of articulation in some transfer conditions (all-pass, LP1600, HP1600, etc). Results: These tests help us to determine the frequency coefficient of the perceptual definition model. Compared with the existing independent probability model and conditional probability model, the new perceptual model has a lesser deviation degree and a better correlation with the measured value.
Key words: speech processing; Chinese speech audiometric; articulation test; speech perception; Chinese monosyllable and disyllabic word
发表期数: 2013年11月第22期
引用格式: 姜旭栋,朱维彬. 汉语单音节和双音节词感知清晰度的测算方法研究[J]. 中国科技论文在线精品论文,2013,6(22):2166-2178.
 
0 评论数 0
暂无评论
友情链接