基于单片机语音识别系统设计
摘要
语音识别是实现人机通信的一种重要技术手段。本文基于AVR RISC结构的单片机，应用隐马尔可夫模型(HMM)理论和方法，完成了针对特定人、小词汇量和孤立词的嵌入式语音识别系统开发。
首先介绍了语音识别过程的各个环节及其实现方法，重点介绍了HMM的基本理论、拓扑结构和算法。通过理论分析和仿真计算完成了系统具体实现环节的设计及参数确定。在上述工作的基础上，进行了识别系统的软件开发、硬件设计和系统调试。最后应用本文设计的系统针对特定人进行了一系列识别实验(包括孤立字、词和短语)。实验结果表明，该系统具有比较好的识别效果。
Speech recognition is an import means of man-machine communication system. An embedded recognition system based on the Hidden Markov Model (HMM) theory is developed with AVR RISC MCU architecture system in this work. The result of a series of experiments shows that the model applied to this system performed very well, could successfully recognize small amount of isolated words of certain speakers with considerable high recognition efficiency.
In this thesis, we firstly give a brief introduction of the speech recognition processes and its realization method, especially of the basic theory, topology structures and algorithms of Hidden Markov Model. The realization method of this system was determined by computer simulation and theory analysis. On the basis of the above work, the hardware and software of the system were designed and tested. In addition, a series of experiments were carried out on this system to recognize isolated numerical, words and phrases of certain speaker in different environments. The results indicate that this speech recognition system has highly recognition efficiency with low cost.
首先介绍了语音识别过程的各个环节及其实现方法，重点介绍了HMM的基本理论、拓扑结构和算法。通过理论分析和仿真计算完成了系统具体实现环节的设计及参数确定。在上述工作的基础上，进行了识别系统的软件开发、硬件设计和系统调试。最后应用本文设计的系统针对特定人进行了一系列识别实验(包括孤立字、词和短语)。实验结果表明，该系统具有比较好的识别效果。
Speech recognition is an import means of man-machine communication system. An embedded recognition system based on the Hidden Markov Model (HMM) theory is developed with AVR RISC MCU architecture system in this work. The result of a series of experiments shows that the model applied to this system performed very well, could successfully recognize small amount of isolated words of certain speakers with considerable high recognition efficiency.
In this thesis, we firstly give a brief introduction of the speech recognition processes and its realization method, especially of the basic theory, topology structures and algorithms of Hidden Markov Model. The realization method of this system was determined by computer simulation and theory analysis. On the basis of the above work, the hardware and software of the system were designed and tested. In addition, a series of experiments were carried out on this system to recognize isolated numerical, words and phrases of certain speaker in different environments. The results indicate that this speech recognition system has highly recognition efficiency with low cost.
引文
[1] 姚天任，《数字语音处理》，华中理工大学出版社，1992．4
[2] 杨行峻、迟惠生，《语音信号数字处理》，电子工业出版社，1995．8
[3] 易可初、田斌、付强，《语音信号处理》，国防工业出版社，2000．6
[4] 盛清，语音自动识别技术及其软件实现，西北工业大学硕士论文，2001．3
[5] 陈方、高升，《语音识别技术及发展》，http://www.36500.com.cn/V0/V3.HTM 2000
[6] 赵大有，语音识别技术，捷通语音识别
[7] 宋叔飚，神经网络在语音识别中的应用研究，西北工业大学硕士论文，2002．3
[8] 耿德根，《AVR高速嵌入式单片机原理与应用》，双龙电子有限公司资料，2000．11
[9] K. N. Stevens, The perception of sounds shaped by resonant circuits, Sc. D. Thesis, Mass. Inst. Tech., Cambridge, Mass., 1952
[10] (美) John G. Ackenhusen, 《实时信号处理》，电子工业出版社，2002．7
[11] Lawrence R. Rabiner, A Tutorial on Hidden Markov Models and Selected Application in Speech Recognition, PROCEEDINGS OF IEEE 1989.2
[12] 刘刚、张洪刚、郭军，用于脱机手写数字识别的隐马尔可夫模型，计算机研究与技术，2003．8
[13] 谢锦辉，线性预测HMM在语音识别、压缩和增强中的应用，华中理工大学博士论文，1990
[14] 何强、毛士艺、张有为，多观察序列连续隐马尔可夫模型的无溢出参数重估，电子学报Vol．28 No．10，2000．10
[15] Baggenstoss P.M., A modified Baum-Welch algorithm for hidden Markov models with multiple observation spaces, Speech and Audio Processing, IEEE Transactions on, Volume: 9, Issue: 4, May 2001
[16] (美) Chi-Tsong chen, 《Digtial Signal Processing Spectral Computation and Filter Design》, 电子工业出版社，2002．9
[17] Wang Chenfa, Lu Chengguo, Sun Linxin, Li Junqing, Speech Enhancement Method for LPC Autoregressive Model and System Implementation of Command Word Recognition used in Noisy Environment, Transactions of Nanjing University of Aeronautics & Astronautics Vol. 15.No.2, Dec. 1998
[18] Renals S., Morgan N., Bourlard H., Cohen M., Franco H., Connectionist probability estimators in HMM speech recognition, Speech and Audio Processing, IEEE Transactions on, Volume: 2, Issue: 1 , Jan. 1994
[19] 石洪波，采用连续混合密度隐马尔可夫模型的语音识别，华北工学院学报，1997．8
[20] 李四信、韦岗，连续型隐马尔可夫模型(HMM)参数与语音识别，华中师范大学学报，1998．3
[21] 张宗红、陈愉、冯晢、邵央、李宗葛，利用HMM嵌入式训练方法建立汉语电话连续语音声学模型，计算机工程与应用，2000．6
[22] 贾宾、朱小燕、罗予频、胡东成，消除溢出问题的精确Baum-Welch算法，软件学报，2000．11
[23] Somervuo P., Speech recognition using temporally connected kernels in mixture density hidden Markov models, Acoustics, Speech, and Signal Processing, 2000. ICASSP '00. Proceedings. 2000 IEEE International Conference on, Volume: 6,5-9 June 2000
[24] Omar M.K., Hasegawa-Johnson M., Levinson S., Gaussian mixture models of phonetic boundaries for speech recognition, Automatic Speech Recognition and Understanding, 2001. ASRU '01. IEEE Workshop on, 9-13 Dec. 2001
[25] 袁俊，HMM连续语音识别中Viterbi算法的优化及应用，电子技术，2001．2
[26] 张杰、黄志同、王晓兰，语音识别中隐马尔可夫模型状态数选取原则及研究，计算机工程与应用，2000．1
[27] ATmega16 datasheet, ATMEL Co., 2003
[28] IS61C512 datasheet, ISSI Co. 1997.9
[29] SST39SF020 datasheet, SST Co. 2003.3
[30] 詹卫前翻译，ImageCraft ICCAVR的中文使用说明，双龙电子有限公司资料，2001
[31] 何强、毛士艺，汉语语音识别的平滑声韵基元HMM算法，北京航空航天大学学报Vol．27 No．2，2001．4
[32] 马明、张杰、王建宇、黄志同，语音识别中隐马尔可夫模型初值估计，数据采集与处理第12卷第2期，1997．6
[33] Huang Xinye, Ma Xiahui, Li Xiang, Fu Yuqing, Lu Jiren, Recognition of Speech Based on HMM/MLP Hybrid Network, Journal of Southeast University(English Edition),Dec.2000 Vol.16 No.2
[34] Gao Hui, Zhou Duqiang, Huang Duansheng, Combined Effect of Noise and Vibration on Articulation of Chinese Syllable, Space Medicine & Medical Engineering Vol.10 No.4, Aug 1997
[35] 郝杰、李星，对经典隐马尔可夫模型的经验性改进，计算机工程与应用，2001．11
[36] Cheok A.D., Chevalier S., Kaynak M., Sengupta K., Ko Chi Chung, Use of a novel generalized fuzzy hidden Markov model for speech recognition, Fuzzy Systems, 2001. The 10th IEEE International Conference on, Volume: 3,2-5 Dec. 2001
[37] Nankaku Y., Tokuda K., Kitamura T., Kobayashi T., Normalized training for HMM-based visual speech recognition, Image Processing, 2000. Proceedings. 2000 International Conference on Volume:
3, 10-13 Sept. 2000
[38] Liu Jianghua, Cheng Junshi, Chen Jiapin, Optimization of HMM Base on Chaos and Algorithm for Hand Gesture Recognition, journal of Systems Engineering and Electronics,Vol. 13, No.4, 2002.4
[39] Mao Xiaoquan, HU Guangrui, Implementation of Grandiment-Based Estimation of HMM Parameters, Journal of Shanghai Jiaotong University, Vol. E-7, No. 1,2002
[40] 李新颖、陆科选，基于HMM模型的语音拨号系统设计，微机与应用，2003年第4期
[41] Sako Shinji, Zen Heiga, Tokuda Keiichi, Masuko Takashi, Kobayashi Takao, Kitamura Tadashi, Improving the performance of HMM-based very low bit rate speech coding, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, v 1, 2003
[42] Ma Gengyu, Lin Xueyin, A HMM based Operational Gesture Recognition Method, Proceedings of SPIE Vol.4552(2001)
[43] 余小清、万旺根、许丽红，基于HMM算法的语音识别系统设计及其混合编程实现，计算机工程 Vol．28 №5 2002．5
[44] 段红梅、汪军、马良河、徐冉，隐马尔可夫模型在语音识别中的应用，工科数学，Vol．18 No．6，2002．12
[2] 杨行峻、迟惠生，《语音信号数字处理》，电子工业出版社，1995．8
[3] 易可初、田斌、付强，《语音信号处理》，国防工业出版社，2000．6
[4] 盛清，语音自动识别技术及其软件实现，西北工业大学硕士论文，2001．3
[5] 陈方、高升，《语音识别技术及发展》，http://www.36500.com.cn/V0/V3.HTM 2000
[6] 赵大有，语音识别技术，捷通语音识别
[7] 宋叔飚，神经网络在语音识别中的应用研究，西北工业大学硕士论文，2002．3
[8] 耿德根，《AVR高速嵌入式单片机原理与应用》，双龙电子有限公司资料，2000．11
[9] K. N. Stevens, The perception of sounds shaped by resonant circuits, Sc. D. Thesis, Mass. Inst. Tech., Cambridge, Mass., 1952
[10] (美) John G. Ackenhusen, 《实时信号处理》，电子工业出版社，2002．7
[11] Lawrence R. Rabiner, A Tutorial on Hidden Markov Models and Selected Application in Speech Recognition, PROCEEDINGS OF IEEE 1989.2
[12] 刘刚、张洪刚、郭军，用于脱机手写数字识别的隐马尔可夫模型，计算机研究与技术，2003．8
[13] 谢锦辉，线性预测HMM在语音识别、压缩和增强中的应用，华中理工大学博士论文，1990
[14] 何强、毛士艺、张有为，多观察序列连续隐马尔可夫模型的无溢出参数重估，电子学报Vol．28 No．10，2000．10
[15] Baggenstoss P.M., A modified Baum-Welch algorithm for hidden Markov models with multiple observation spaces, Speech and Audio Processing, IEEE Transactions on, Volume: 9, Issue: 4, May 2001
[16] (美) Chi-Tsong chen, 《Digtial Signal Processing Spectral Computation and Filter Design》, 电子工业出版社，2002．9
[17] Wang Chenfa, Lu Chengguo, Sun Linxin, Li Junqing, Speech Enhancement Method for LPC Autoregressive Model and System Implementation of Command Word Recognition used in Noisy Environment, Transactions of Nanjing University of Aeronautics & Astronautics Vol. 15.No.2, Dec. 1998
[18] Renals S., Morgan N., Bourlard H., Cohen M., Franco H., Connectionist probability estimators in HMM speech recognition, Speech and Audio Processing, IEEE Transactions on, Volume: 2, Issue: 1 , Jan. 1994
[19] 石洪波，采用连续混合密度隐马尔可夫模型的语音识别，华北工学院学报，1997．8
[20] 李四信、韦岗，连续型隐马尔可夫模型(HMM)参数与语音识别，华中师范大学学报，1998．3
[21] 张宗红、陈愉、冯晢、邵央、李宗葛，利用HMM嵌入式训练方法建立汉语电话连续语音声学模型，计算机工程与应用，2000．6
[22] 贾宾、朱小燕、罗予频、胡东成，消除溢出问题的精确Baum-Welch算法，软件学报，2000．11
[23] Somervuo P., Speech recognition using temporally connected kernels in mixture density hidden Markov models, Acoustics, Speech, and Signal Processing, 2000. ICASSP '00. Proceedings. 2000 IEEE International Conference on, Volume: 6,5-9 June 2000
[24] Omar M.K., Hasegawa-Johnson M., Levinson S., Gaussian mixture models of phonetic boundaries for speech recognition, Automatic Speech Recognition and Understanding, 2001. ASRU '01. IEEE Workshop on, 9-13 Dec. 2001
[25] 袁俊，HMM连续语音识别中Viterbi算法的优化及应用，电子技术，2001．2
[26] 张杰、黄志同、王晓兰，语音识别中隐马尔可夫模型状态数选取原则及研究，计算机工程与应用，2000．1
[27] ATmega16 datasheet, ATMEL Co., 2003
[28] IS61C512 datasheet, ISSI Co. 1997.9
[29] SST39SF020 datasheet, SST Co. 2003.3
[30] 詹卫前翻译，ImageCraft ICCAVR的中文使用说明，双龙电子有限公司资料，2001
[31] 何强、毛士艺，汉语语音识别的平滑声韵基元HMM算法，北京航空航天大学学报Vol．27 No．2，2001．4
[32] 马明、张杰、王建宇、黄志同，语音识别中隐马尔可夫模型初值估计，数据采集与处理第12卷第2期，1997．6
[33] Huang Xinye, Ma Xiahui, Li Xiang, Fu Yuqing, Lu Jiren, Recognition of Speech Based on HMM/MLP Hybrid Network, Journal of Southeast University(English Edition),Dec.2000 Vol.16 No.2
[34] Gao Hui, Zhou Duqiang, Huang Duansheng, Combined Effect of Noise and Vibration on Articulation of Chinese Syllable, Space Medicine & Medical Engineering Vol.10 No.4, Aug 1997
[35] 郝杰、李星，对经典隐马尔可夫模型的经验性改进，计算机工程与应用，2001．11
[36] Cheok A.D., Chevalier S., Kaynak M., Sengupta K., Ko Chi Chung, Use of a novel generalized fuzzy hidden Markov model for speech recognition, Fuzzy Systems, 2001. The 10th IEEE International Conference on, Volume: 3,2-5 Dec. 2001
[37] Nankaku Y., Tokuda K., Kitamura T., Kobayashi T., Normalized training for HMM-based visual speech recognition, Image Processing, 2000. Proceedings. 2000 International Conference on Volume:
3, 10-13 Sept. 2000
[38] Liu Jianghua, Cheng Junshi, Chen Jiapin, Optimization of HMM Base on Chaos and Algorithm for Hand Gesture Recognition, journal of Systems Engineering and Electronics,Vol. 13, No.4, 2002.4
[39] Mao Xiaoquan, HU Guangrui, Implementation of Grandiment-Based Estimation of HMM Parameters, Journal of Shanghai Jiaotong University, Vol. E-7, No. 1,2002
[40] 李新颖、陆科选，基于HMM模型的语音拨号系统设计，微机与应用，2003年第4期
[41] Sako Shinji, Zen Heiga, Tokuda Keiichi, Masuko Takashi, Kobayashi Takao, Kitamura Tadashi, Improving the performance of HMM-based very low bit rate speech coding, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, v 1, 2003
[42] Ma Gengyu, Lin Xueyin, A HMM based Operational Gesture Recognition Method, Proceedings of SPIE Vol.4552(2001)
[43] 余小清、万旺根、许丽红，基于HMM算法的语音识别系统设计及其混合编程实现，计算机工程 Vol．28 №5 2002．5
[44] 段红梅、汪军、马良河、徐冉，隐马尔可夫模型在语音识别中的应用，工科数学，Vol．18 No．6，2002．12