A feedforward neural network learning algorithm based on multi-dimensional unit sphere sum logic is proposed and successfully applied to the path planning of mobile robots and the expert system for source rock evaluation. This research work won the second prize of Lin Jiaqiao Applied Mathematics. In recent years, acoustic modeling, fast decoding algorithm, language model, prosody model, etc. Large vocabulary Chinese continuous speech recognition based on segmented model: Corpus-based splicing method, HMM-based parameter synthesis, duration modeling and so on. Used for Chinese speech synthesis; Speech enhancement based on multi-statistical distribution noise estimation and acoustic mechanism filter design, as well as multi-pitch estimation based on statistics and organizational strategy based on auditory mechanism for computational auditory scene analysis have been deeply studied, and many important scientific research results have been achieved. Currently applying for a patent. Combined with scientific research tasks, it has guided and trained dozens of doctoral and master students. The National Natural Science Foundation and the 863 High-tech Development Plan are mainly completed as follows:

1, "Anti-noise speech recognition method based on segment feature distribution and auditory mechanism", supported by National Natural Science Foundation (No.:60172055);

2. "Auditory computing model and its application in speaker recognition", a key project of Natural Science Foundation (cooperation with Peking University) (No.:69635020);

3. "863" project in the Ninth Five-Year Plan (No.:863-306-ZT06-01-03);

4. "Research on integrated voice navigation GIS/GPS system and its implementation in PDA", supported by Beijing Natural Science Foundation (No.:4002012);

5. The frontier project in the field of knowledge innovation engineering of China Academy of Sciences "Establishment of Chinese Universal Speech Database and Data Conversion Algorithm" (No.:1m02j05);

6. "Research on speaker confirmation technology of confirmation text and its application in financial field" (in cooperation with Peking University), Tenth Five-Year "863" project (No.:200114181);

7. "Research on Key Technologies of Personalized Chinese Speech Synthesis Method", supported by Beijing Natural Science Foundation (No.:4042025).

He published more than 100 academic papers, some of which were published in recent years:

1, Liu Wenju, Sun Bing, Zhong Qiuhai, "Research on Hierarchical Speaker Recognition Based on Speaker Classification", Journal of Electronic Science, Vol.33, No.7, pp 1230~ 1233, 2005.

2,,, Zhang,, "A Fast Segmentation Model Framework for Avoiding Redundant Computation of Segmentation", International Journal of Computational Linguistics and Chinese Processing, vol. 1 1, pp. 73-86, March 2006.

3. Tang Yun, Liu Wenju, zhanghua, Xu Bo, Ding Guohong, "A Piecewise Speech Decoding Algorithm from Coarse to Fine", Volume I of Proceedings of ICASSP (International Conference on Acoustics, Speech and Signal Processing), pp44 1~444, May 200614-/kloc-0.

4. Kang Heng and Liu Wenju, "Direct spectral representation based on selective linear predictive coding and its application in spectral smoothing", Journal of China Geo University. Interspec 2006-ICSLP (9th International Conference on Oral Processing), pp2050~2053, September 2006 17-2 1, Pittsburgh, Pennsylvania, USA.

5. Peng Li, Guan Yong, Liu Wenju and Xu Bo, "Monoear Speech Separation and Speech Objective Quality Assessment Based on Computational Auditory Scene Analysis", IEEE Transactions on Audio, Speech and Language Processing, Volume 14, No.6, pp20 14~2023, 20061/.

6. Peng Shouye, "Fast and Robust Random Segmentation Model for Chinese Numeric String Recognition", 2008 IEEE Conference on Computing Intelligence (WCCI 2008), pp. 1446~ 1450, June 2008 1-6, China.

7. Cheng Ning, Liu Wenju, Peng Li, Xu Bo, "Effective Microphone Array Post-filter in Any Environment", Proceedings of the 9th Annual Conference of International Speech Communication Association (2008), pp. 439-442, 22-26 September 2008, Brisbane, Australia.

Peng Shouye, "Research on Chinese speech synthesis based on segmented acoustic model", Proceedings of the 6th International Symposium on Neural Networks, pp. 928-935, 2009, May 26-29, 2009.

9. Zhang Xueliang, Liu Wenju, Peng Li and Xu Bo, "Mono voiced speech separation based on fine harmonic grouping strategy", Proceedings of ICASSP'2009(IEEE International Conference on Acoustics, Speech and Signal Processing), pp. 466 1~4664, April124, 2009, Taipei.

10, Peng Li, Guan Yong, Wang Shijin, Xu Bo, Liu Wenju, "Robust speech recognition of monophonic speech separation based on MAXVQ and CASA", Computer Speech and Language 24 (20 10), pp. 30-44.