符合人耳听觉特性的语音音质的客观评价方法  

Objective Assessment of Acoustic Fidelity of Speech Based on Characteristics of Human Hearing

作　　者：陈静[1] 周毅刚[1] 周建林[1] 

机构地区：[1]哈尔滨工业大学航天电子与光电工程系,150001

出　　处：《哈尔滨工业大学学报》1998年第6期80-83,共4页Journal of Harbin Institute of Technology

摘　　要：对于合成语音的音质进行评价，是评价合成系统使用价值和性能的一种主要方式．通过音质评价，找出影响音质的原因，反过来指导音质改善，这正好体现了人类发声时听觉的反馈作用．现行的评价方案多为主观评价，不仅不利操作，而且不能对改善音质提供定量的建议．为了实现对语音音质的客观评价，作者通过对LSP参数的频谱敏感度的研究，将其与听觉特性相联系，用频谱敏感度表征语音中对听觉敏感的成分，从而给出了符合人耳听觉特性的音质的定量描述，提供了一种对语音音质的客观评价方法．并通过实验验证了这种方法的正确性．The assessment of acoustic fidelity of synthetic speech is a main way of assessing the performance andfonction of a synthesis system. To find the causes of and to improve the speech of worse acoustic fidelity is just likethe feedback effect of hearing. The way of assessment now used generally is a subjective one which can not give aquantitative analysis for improving and it is quite complicated. With human hearing characteristics considered together with the sensitivity of LSP parameters to their spectral envelope, the acoustic fidelity of synthetic CHINESEspeech is expressed in a quanitative way, and an objective assessment method for acoustic fidelity of synthetic CHINESE speech is thus proposed.

关 键 词：LSP 汉语语音 听觉特性 合成语音 音质 评价 

分 类 号：TN912.3[电子电信—通信与信息系统]