Tsinghua University|中文版

Speech and Language Technology R&D Center gained support from national key research and development plan

Recently, 2017 national key research and development project “Research on Conformance testing technology of intelligent speech products” has got approved. The project was led by the third institute of Ministry of Public Security and the fifth topic “Key quality detection technology of voiceprint recognition products” was jointly undertaken by Speech and Language Technology R&D Center (Center for short)of RIIT and Beijing D-Ear Technologies Co., Ltd. The team has gathered a total of twelve institutions in China, including smart speech research institutes, colleges and universities, and industrial application enterprises.

Center is responsible for three key tasks in this project: firstly, study the core factors of influencing voiceprint recognition performance and application effects and analyze the conditions, mechanism and relationship of their impacts; secondly, study and design the anti-noise performance parameters, time-varying performance parameters and anti-attack intrusion performance parameters of voiceprint recognition; thirdly, to achieve the typical voiceprint application security of mobile devices and digital dynamic password, we study key features in accordance with refined products, analyze whether the evaluation index system is scientific and reasonable, and require key performance indicators covering voiceprint recognition and the current technical indicators.


YK)Q]G6ZP_ABA4~YT8_FN3P.png

声纹时变鲁棒性研究

574X1AXS4U827H037U9C]`A.png

防声纹攻击闯入研究

7(A$W}OU}II}8Y5F1@E(A~8.png

移动设备和动态数字密码的典型声纹应用场景

As an emerging technology with strategic significance, intelligent speech technology has long been the focus of science industry with all countries striving to achieve technological breakthroughs. Center of RIIT has seized the opportunities and made innovations and pioneering achievements in the fields of speech recognition and voiceprint recognition in recent years. The novel extraction method has been proposed based on speakers’ features which can greatly promote the application of the recognition of speakers’ performance and playback detection. Center and Beijing D-Ear Technologies Co., Ltd. established a joint laboratory of voiceprint processing and applied the scientific research results to the service industry. The product “sound security” launched by Beijing D-Ear Technologies Co., Ltd. has been used by China Construction Bank and many other enterprises.

 

【2017-07-07】【Read:20】