Proposed a feedforward neuron network learning algorithm based on multi-dimensional unit sphere and logic, and successfully applied it to mobile robot path planning and oil-generating rock evaluation expert system. This research work was awarded Lin Jiaqiao's second degree in Applied Mathematics. Prizes. In recent years, acoustic modeling, fast decoding algorithms, language models, prosody models, etc. for large vocabulary Chinese continuous speech recognition based on segment models; corpus splicing methods for Chinese speech synthesis, HMM-based parameter synthesis, duration modeling, etc. ; We have conducted in-depth research on filter design based on multi-statistical distribution noise estimation and acoustic mechanism for speech enhancement, as well as multi-pitch estimation based on statistics and organization strategies based on auditory mechanism for computational auditory scene analysis, and have achieved a number of important scientific research results. . The patent is currently pending. In combination with scientific research tasks, he has guided and trained dozens of doctoral and master's students. The main completed National Natural Science Foundation and 863 High-Tech Development Plan are as follows:
1. "Anti-noise speech recognition method based on segment feature distribution and auditory mechanism", National Natural Science Foundation project (No. 60172055 );
2. "Auditory Computational Model and Its Application in Speaker Recognition," (in cooperation with Peking University) Natural Science Foundation Key Project (No. 69635020);
3. "Research on Practical and Fast Algorithms for NP Problems in Markov Class Models" Ninth Five-Year Plan "863" Project (No. 863-306-ZT06-01-03);
4. "Voice Navigation Research on GIS/GPS integrated system and its implementation on PDA", Beijing Natural Science Foundation Project (No. 4002012);
5. "Establishment of Chinese universal speech database and data conversion algorithm", Chinese Academy of Sciences Frontier Project in the Field of Knowledge Innovation Engineering (No.: 1M02J05);
6. "Research on Limited Text Speaker Confirmation Technology and Its Application in the Financial Field," (in cooperation with Peking University), 15th "863 "Project (No.: 2001AA114181);
7. "Research on Key Technologies of Voice Personalized Chinese Speech Synthesis Method," Beijing Natural Science Foundation Project (No.: 4042025).
Published more than 100 academic papers, some papers published in recent years:
1. Liu Wenju, Sun Bing, Zhong Qiuhai, "Research on hierarchical speaker recognition based on speaker classification technology," Journal of Electronics, Volume 33, Issue 7, pp1230~1233, 2005.
2. Yun Tang, Wenju Liu, Yiyan Zhang, Bo Xu, "A Framework for Fast Segment Model by Avoidance of Redundant Computation on Segment," International Journal of Computational Linguistics and Chinese Language Processing, Vol. 11 (1), pp73~86, Mar., 2006.
3. Yun Tang, Wen-Ju Liu, Hua Zhang, Bo Xu, Guo-Hong Ding, “One-pass Coarse-to-Fine Segmental Speech Decoding Algorithm,” Proceedings of ICASSP (International Conference on Acoustics, Speech, and Signal Processing, ) Vol.Ⅰ, pp441~444, May 14-19, 2006, Toulouse, France.
4. Kang Heng, and Liu Wenju, "Selective-LPC based Representation of STRAIGHT Spectrum and Its Applications in Spectral Smoothing," Proc. of INTERSPEECH 2006-ICSLP(9th International Conference on Spoken Language Processing), pp2050~2053, September 17-21, 2006, Pittsburgh, Pennsylvania, USA.
5. Peng Li, Yong Guan, Wenju Liu and Bo Xu, "Monaural Speech Separation Based on Computational Auditory Scene Analysis and Objective Quality Assessment of Speech," IEEE Transaction on Audio, Speech, and Language Processing, Vol.14, Num.6, pp2014~2023, November, 2006.
6. Wenju Liu, Yun Tang, and Shouye Peng, "Fast and Robust Stochastic Segment Model for Mandarin Digital String Recognition,” Proceedings of 2008 IEEE Congress on Computational Intelligence (WCCI 2008), pp.1446~1450, June 1-6, 2008, Hong Kong, China.
7. Ning Cheng, Wen-Ju Liu, Peng Li, and Bo Xu, “An Effective Microphone Array Post-filter in Arbitrary Environments,” Proceedings of 9th Annual Confere
nce of the International Speech Communication Association (Interspeech 2008), pp.439~442, 22-26 September 2008, Brisbane, Australia.
8. Wenju Liu, Yun Tang, and Shouye Peng, “Research on Segment Acoustic Model Based Mandarin LVCSR,” Proceedings of 6th International Symposium on Neural Networks, pp.928~935, ISNN 2009 Wuhan, China, May 26-29, 2009.
9. Xueliang Zhang, Wenju Liu , Peng Li, and Bo Xu, "MONAURAL VOICED SPEECH SEGREGATION BASED ON ELABORATE HARMONIC GROUPING STRATEGY," Proceedings of ICASSP'2009 (IEEE International Conference on Acoustics, Speech, and Signal Processing), pp.4661~4664, April 19~24 , 2009, Taipei International Convention Center, Taipei, Taiwan.
10. Peng Li, Yong Guan, Shijin Wang, Bo Xu and Wenju Liu, “Monaural Speech Separation Based on MAXVQ and CASA for Robust Speech Recognition, ” Computer Speech and Language 24 (2010) pp. 30~44.