项目名称: 语音信息保护中的声学掩蔽研究
项目编号: No.11474307
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 数理科学和化学
项目作者: 许勇
作者单位: 中国科学院声学研究所
项目金额: 80万元
中文摘要: 为解决日常语言交流中语音易被窃听、盗录的问题,本项目围绕语音信息保护中的声学掩蔽开展研究。项目以对人耳听觉感知机理、传声器拾音机理和语音信号处理理论的深入研究为基础,通过理论、实验研究相结合的方式,重点探索并解决:1,超声对MEMS传声器接收性能的影响规律;2,信号处理算法对低信噪比语音的言语可懂度改善原理;3,低可懂度语音的可懂度主客观评价方法的科学问题。进而提出一系列超声、可听声掩蔽信号设计方法,以降低目标语音可懂度,实现信息保护的目的。 本项目首次在考虑传声器拾音和语音信号处理环节对言语可懂度影响的前提下开展声掩蔽研究,提出的声掩蔽方法具有掩蔽信号能量小、烦恼度低、能对抗可懂度增强算法的优点,可有效实现对目标语音信息的保护。项目工作对深入理解语音的信息掩蔽效应和影响言语可懂度的各种声学特征等关键因素都有重要意义,同时可为保护语音信息安全的后续应用研究提供理论依据,具有实际应用价值。
中文关键词: 声掩蔽;言语可懂度;语音信号处理;听觉感知
英文摘要: Speech is one of the most important forms of communication. The issue of privacy of voice has been dealt with as a research challenge. In order to protect the information of the conversation, the sound masking theory and method will be studied in this project. Firstly, the theory of the auditory perception, the mechanism of MEMS microphone working and the theory of speech signal processing will be investigated comprehensively. Secondly, the various factors that influence the masking effect will be considered and analyzed. Different kinds of sound masking signals will be designed in audio frequency and ultrasonic frequency, respectively. The subjective and objective experiments will be carried on to verify the theory work. Compared to the previous research work, the novel sound masking solution proposed by this project is much effective for depressing the speech intelligibility. The advantages can be summarized as follows: The annoyance level of our sound masking signal will be lower than noise signal's. Meanwhile, the current speech enhancement algorithms can hardly improve the intelligibility of the speech mixed with our masking signal. The work of this project is helpful to understand the important characteristics of speech intelligibility and the effect of sound masking. The productions of the project could be used in applications of private speech information protection.
英文关键词: Sound Masking;Speech Intelligibility;Speech Signal Processing;Auditory Perception