轮椅机器人语音识别控制系统的研究与实现
摘要
机器人发展至今,智能机器人成为研究的热点之一。对于机器人的控制,语音控制无非是最自然、最便捷的控制方式。语音识别技术使机器人能听懂人的自然语言,由识别得到的信息可以作为声控信号应用到机器人的多种技术领域。智能轮椅作为助老/助残服务机器人系列产品中的一个重要研究领域,将语音识别技术应用于普通轮椅,用语音来控制轮椅的运动为使用者带来了极大的方便。因此研究并开发实用的机器人语音识别控制系统对于智能机器人的广泛应用具有重要的意义。论文的主要内容如下;
首先,实现了轮椅机器人语音指令识别算法。基于语音识别的原理,对轮椅语音指令进行了预滤波、数字化、预加重处理、分帧、小波降噪和双门限端点检测;对Mel频率倒谱系数(MFCC)与线性预测倒谱系数(LPCC)的性能进行了对比分析,选取性能较好的MFCC作为识别特征参数,采用动态时间规整(DTW)的高效算法,实现了易于硬件实现、识别性能较好、执行效率较高的轮椅机器人语音指令识别算法。
其次,设计完成了轮椅机器人的语音识别控制系统。采用凌阳16位单片机SPCE061A作为系统核心,完成了轮椅机器人语音识别控制系统的硬件设计;基于SPCE061A的集成开发环境编写了轮椅语音识别控制系统的软件。实验结果表明,实现的语音识别控制系统的识别性能及轮椅运动的机械性能较好。同时,该系统结构简单,性价比高,易于功能扩展和移植,具有广阔的应用前景。
With the development of robot, the intelligent robot becomes one research focus. For the control of the robot, speech control is the most natural and convenient way. Speech recognition technology can make robot understand human language, the signal from recognition can be used as voice control signal and applied in many technology domains of robot. The intelligent wheelchair is one important research domain of the series of the elder-aid and handicapped-aid robots. By applying speech recognition technology to the common wheelchair, it is very convenient to use voice to control the movement of wheelchair. Therefore, it is very important to research and develop practical robot speech recognition control system in order to apply the service robot widely. The main content of the paper is below.
Firstly, the speech instruction recognition algorithm of the wheelchair robot is realized. According to the theory of speech recognition, the speech instruction of the wheelchair is pre-processed, including pre-filtering, digitalization, pre-emphasis, packetization, wavelet denoising, endpoint detection with double threshold. The performance of mel frequency cepstral coefficients and linear predictive cepstral coefficients is contrasted and analyzed, MFCC has better performance and it is selected as the feature parameter. The more efficient dynamic time warping algorithm is adopted as the recognition algorithm. The wheelchair instruction recognition algorithm is programmed, it has better performance and efficiency and it is easily realized on hardware.
Secondly, the speech recognition control system of the wheelchair robot is designed. The hardware design of the speech recognition control system which selects SPCE061A as the core is completed, and the software of the system is programmed in the integrated developing environment of SPCE061A. The experiment result shows that the system has better recognition performance and mechanical performance. In the meantime, the system has simple structure and better performance-price ratio, and it can be enlarged in function and transplanted easily. Therefore, the speech recognition system has broad application prospects.
首先,实现了轮椅机器人语音指令识别算法。基于语音识别的原理,对轮椅语音指令进行了预滤波、数字化、预加重处理、分帧、小波降噪和双门限端点检测;对Mel频率倒谱系数(MFCC)与线性预测倒谱系数(LPCC)的性能进行了对比分析,选取性能较好的MFCC作为识别特征参数,采用动态时间规整(DTW)的高效算法,实现了易于硬件实现、识别性能较好、执行效率较高的轮椅机器人语音指令识别算法。
其次,设计完成了轮椅机器人的语音识别控制系统。采用凌阳16位单片机SPCE061A作为系统核心,完成了轮椅机器人语音识别控制系统的硬件设计;基于SPCE061A的集成开发环境编写了轮椅语音识别控制系统的软件。实验结果表明,实现的语音识别控制系统的识别性能及轮椅运动的机械性能较好。同时,该系统结构简单,性价比高,易于功能扩展和移植,具有广阔的应用前景。
With the development of robot, the intelligent robot becomes one research focus. For the control of the robot, speech control is the most natural and convenient way. Speech recognition technology can make robot understand human language, the signal from recognition can be used as voice control signal and applied in many technology domains of robot. The intelligent wheelchair is one important research domain of the series of the elder-aid and handicapped-aid robots. By applying speech recognition technology to the common wheelchair, it is very convenient to use voice to control the movement of wheelchair. Therefore, it is very important to research and develop practical robot speech recognition control system in order to apply the service robot widely. The main content of the paper is below.
Firstly, the speech instruction recognition algorithm of the wheelchair robot is realized. According to the theory of speech recognition, the speech instruction of the wheelchair is pre-processed, including pre-filtering, digitalization, pre-emphasis, packetization, wavelet denoising, endpoint detection with double threshold. The performance of mel frequency cepstral coefficients and linear predictive cepstral coefficients is contrasted and analyzed, MFCC has better performance and it is selected as the feature parameter. The more efficient dynamic time warping algorithm is adopted as the recognition algorithm. The wheelchair instruction recognition algorithm is programmed, it has better performance and efficiency and it is easily realized on hardware.
Secondly, the speech recognition control system of the wheelchair robot is designed. The hardware design of the speech recognition control system which selects SPCE061A as the core is completed, and the software of the system is programmed in the integrated developing environment of SPCE061A. The experiment result shows that the system has better recognition performance and mechanical performance. In the meantime, the system has simple structure and better performance-price ratio, and it can be enlarged in function and transplanted easily. Therefore, the speech recognition system has broad application prospects.
引文
[1]韩纪庆,张磊,郑铁然.语音信号处理.北京;清华大学出版社,2004
[2]王炳锡,屈丹,彭煊.实用语音识别基础.北京;国防工业出版社,2004
[3]罗志增,赵敬斌.机器人语音控制及其实现.杭州电子工业学院学报,2004,24(1);30-34
[4]诸刚.汉语语音识别技术在机器人控制中的应用.北京市计划劳动管理干部学院学报,2004,12(1);47-48
[5]何清华,陈阳,刘永强.智能轮椅中人机界面的分类及评估.现代电子技术,2003(17);20-23
[6]李晶皎.嵌入式语音技术及凌阳16位单片机应用.北京;北京航空航天出版社,2003
[7]王富中,黄文浩.基于语音识别技术的智能控制系统设计.自动化与仪器,2006(4);8-10
[8]杨尚国,杨金龙.语音识别技术概述.福建电脑,2006(8);34-34
[9]陈立万.基于语音识别系统中DTW算法改进技术研究.微计算机信息,2006,22(2);267-268
[10]荆嘉敏,刘加,刘润生.基于HMM的语音识别技术在嵌入式系统中的应刚.电子技术应用,2003(10);12-14
[11]赵力.语音信号处理.北京;机械工业出版社,2003
[12]Xiang Xie,Junhui Zhao,Jingming Kuang.Comparative study on VQ-Based efficient mandarin speech recognition method.Journal of Beijing Institute of Technology,2002,Vol.11(3);266-270
[13]Kundu,A.(U.S.West Advanced Technologies),Bayya,A.Speech recognition using hybrid hidden Markov model and NN classifier.International Journal of Speech Technology,1998,Vol.2(3);227-240
[14]田岚,陆小珊,白树忠.基于快速神经网络算法的非特定人语音识别.控制与决策,2002,17(1);65-68
[15]徐秀平,李柱峰.基于神经网络的自学习非特定人语音识别研究.电声技术,2004(6);30-32
[16]刘加.汉语大词汇量连续语音识别系统研究进展.电子学报,2000,28(1);85-90
[17]何湘智.语音识别的研究与进展.计算机与现代化,2002(3);3-6
[18]郭晓丹.计算机语音识别技术问题漫谈.自动化信息,2005,49(5);43-45
[19]Jie Li,Jiaxin Wang,Yannan Zhao and Zehong Yang.Self-adaptive design of hidden Markov models.SHORT COMMUNICATION.Pattern Recognition Letter,2004,Vol.25(2);197-210
[20]刘放军,王仁华.语音识别前端鲁棒性问题综述.计算机科学,2006,33(4);168-173
[21]诸刚.用于机器人控制的汉语语音识别系统的研究与实现;[硕士学位论文].天津; 天津大学,2004
[22]周伟.MATLAB小波分析高级技术.西安;西安电子科技大学出版社,2005
[23]飞思科技产品研发中心.小波分析理论与MATLAB7实现.北京;电子工业出版社,2005
[24]许山川.基于小波变换的语音信号去噪研究;[硕士学位论文].秦皇岛;燕山大学,2006
[25]文莉,刘正士,葛远建.小波去噪的几种方法.合肥工业大学学报(自然科学版),2002,25(2);42-47
[26]哈恒旭,桑在中.小波分析在数字滤波中的应用研究.电力自动化设备,1999,19(65);12-16
[27]Xu B Z,Qiu W(1992).A new method of Chinese consonants recognition based on time-domain features and phonetic knowledge.ICIIPS'92;515-518
[28]刘倩.语音识别技术在控制系统中的应用研究;[硕士学位论文].沈阳;东北大学,2006
[29]易克初,田斌,付强.语音信号处理.北京;国防工业出版社,2000
[30]K.Yao,E.Shi,etc.Residual noise compensation for robust speech recognition non-stationary noise.Proc.ICASSP,2000(2);1125-1128
[31]张雄伟,陈亮等.现代语音处理技术及应用.北京;机械工业出版社,2003
[32]马俊.语音识别技术研究;[硕士学位论文].哈尔滨;哈尔滨工程大学,2004
[33]夏敏磊.语音端点检测技术研究;[硕士学位论文].杭州;浙江大学,2005
[34]Tze Fen Li,Shui-Ching Chang and Chung-Bow Lee.A simple statistical speech recognition of mandarin monosyllables.Applied Mathematics and Computation,2006,Vol.177(2);644-651
[35]马昕,杜利民.基于小波调制尺度的语音特征参数提取方法.计算机应用,2005,26(6);1342-1344
[36]Gowdy,J.N.(Clemson Univ); Tufekci,Z.Mel-scaled discrete wavelet coefficients for speech recognition.ICASSP,IEEE International Conference on Acoustics,Speech and Signal Processing-Proceedings,2000(3);1351-1354
[37]Lavner Y,Gath I,Rosenhouse J.The effects of acoustic modifications on the identification of familiar voices speaking isolated vowels.Speech Communication,2000,Vol.30(1);9-26
[38]王让定,柴佩琪.语音倒谱特征的研究.计算机工程,2003,29(13);31-33
[39]何强,何英.MATLAB扩展编程.北京;清华大学出版社,2002
[40]金晶.轮椅机器人语音识别、控制技术的研究;[硕士学位论文].苏州;苏州大学,2006
[41]魏力.嵌入式语音识别系统的研究;[硕士学位论文].武汉;武汉理工大学,2006
[42]崔健.语音识别技术的研究及应用;[硕士学位论文].沈阳;东北大学,2006
[43]蔡莲红,黄德智.现代语音技术基础与应用.北京;清华大学出版社,2003
[44]张有为,甘俊英,何强等.人机自然交互.北京;清华大学出版社,2003
[45]罗亚非.凌阳16位单片机应用基础.北京;国防工业出版社,2004
[46]宋超,姜力,赵大威.基于SPCE061A的多自由度假手语音控制系统的研究.机械与电 子,2006(10);40-42
[47]张培仁,张志坚,高修峰.十六位单片机微处理器原理及应用(凌阳SPCE061A).北京;清华大学出版社,2005
[48]颜利彪,范蟠果.基于单片机的简易智能电动车.电子技术,2004(4);8-10
[49]叶聪红,徐文龙,王效杰等.两轮驱动小车系统的设计实现及其路径规划.航天制造技术,2005(5);15-21
[50]王晓明.电动机的单片机控制.北京;北京航空航天大学出版社,2002
[51]侯媛彬,袁益民,崔汉平.凌阳单片机原理及其毕业设计精选.北京;科学出版社,2006
[52]凌阳科技教育推广中心.SPCE061A在语音识别机器人中的应用.http;//www.unsp.com/app/html/2006424100102.shtml,2006-4-24
[2]王炳锡,屈丹,彭煊.实用语音识别基础.北京;国防工业出版社,2004
[3]罗志增,赵敬斌.机器人语音控制及其实现.杭州电子工业学院学报,2004,24(1);30-34
[4]诸刚.汉语语音识别技术在机器人控制中的应用.北京市计划劳动管理干部学院学报,2004,12(1);47-48
[5]何清华,陈阳,刘永强.智能轮椅中人机界面的分类及评估.现代电子技术,2003(17);20-23
[6]李晶皎.嵌入式语音技术及凌阳16位单片机应用.北京;北京航空航天出版社,2003
[7]王富中,黄文浩.基于语音识别技术的智能控制系统设计.自动化与仪器,2006(4);8-10
[8]杨尚国,杨金龙.语音识别技术概述.福建电脑,2006(8);34-34
[9]陈立万.基于语音识别系统中DTW算法改进技术研究.微计算机信息,2006,22(2);267-268
[10]荆嘉敏,刘加,刘润生.基于HMM的语音识别技术在嵌入式系统中的应刚.电子技术应用,2003(10);12-14
[11]赵力.语音信号处理.北京;机械工业出版社,2003
[12]Xiang Xie,Junhui Zhao,Jingming Kuang.Comparative study on VQ-Based efficient mandarin speech recognition method.Journal of Beijing Institute of Technology,2002,Vol.11(3);266-270
[13]Kundu,A.(U.S.West Advanced Technologies),Bayya,A.Speech recognition using hybrid hidden Markov model and NN classifier.International Journal of Speech Technology,1998,Vol.2(3);227-240
[14]田岚,陆小珊,白树忠.基于快速神经网络算法的非特定人语音识别.控制与决策,2002,17(1);65-68
[15]徐秀平,李柱峰.基于神经网络的自学习非特定人语音识别研究.电声技术,2004(6);30-32
[16]刘加.汉语大词汇量连续语音识别系统研究进展.电子学报,2000,28(1);85-90
[17]何湘智.语音识别的研究与进展.计算机与现代化,2002(3);3-6
[18]郭晓丹.计算机语音识别技术问题漫谈.自动化信息,2005,49(5);43-45
[19]Jie Li,Jiaxin Wang,Yannan Zhao and Zehong Yang.Self-adaptive design of hidden Markov models.SHORT COMMUNICATION.Pattern Recognition Letter,2004,Vol.25(2);197-210
[20]刘放军,王仁华.语音识别前端鲁棒性问题综述.计算机科学,2006,33(4);168-173
[21]诸刚.用于机器人控制的汉语语音识别系统的研究与实现;[硕士学位论文].天津; 天津大学,2004
[22]周伟.MATLAB小波分析高级技术.西安;西安电子科技大学出版社,2005
[23]飞思科技产品研发中心.小波分析理论与MATLAB7实现.北京;电子工业出版社,2005
[24]许山川.基于小波变换的语音信号去噪研究;[硕士学位论文].秦皇岛;燕山大学,2006
[25]文莉,刘正士,葛远建.小波去噪的几种方法.合肥工业大学学报(自然科学版),2002,25(2);42-47
[26]哈恒旭,桑在中.小波分析在数字滤波中的应用研究.电力自动化设备,1999,19(65);12-16
[27]Xu B Z,Qiu W(1992).A new method of Chinese consonants recognition based on time-domain features and phonetic knowledge.ICIIPS'92;515-518
[28]刘倩.语音识别技术在控制系统中的应用研究;[硕士学位论文].沈阳;东北大学,2006
[29]易克初,田斌,付强.语音信号处理.北京;国防工业出版社,2000
[30]K.Yao,E.Shi,etc.Residual noise compensation for robust speech recognition non-stationary noise.Proc.ICASSP,2000(2);1125-1128
[31]张雄伟,陈亮等.现代语音处理技术及应用.北京;机械工业出版社,2003
[32]马俊.语音识别技术研究;[硕士学位论文].哈尔滨;哈尔滨工程大学,2004
[33]夏敏磊.语音端点检测技术研究;[硕士学位论文].杭州;浙江大学,2005
[34]Tze Fen Li,Shui-Ching Chang and Chung-Bow Lee.A simple statistical speech recognition of mandarin monosyllables.Applied Mathematics and Computation,2006,Vol.177(2);644-651
[35]马昕,杜利民.基于小波调制尺度的语音特征参数提取方法.计算机应用,2005,26(6);1342-1344
[36]Gowdy,J.N.(Clemson Univ); Tufekci,Z.Mel-scaled discrete wavelet coefficients for speech recognition.ICASSP,IEEE International Conference on Acoustics,Speech and Signal Processing-Proceedings,2000(3);1351-1354
[37]Lavner Y,Gath I,Rosenhouse J.The effects of acoustic modifications on the identification of familiar voices speaking isolated vowels.Speech Communication,2000,Vol.30(1);9-26
[38]王让定,柴佩琪.语音倒谱特征的研究.计算机工程,2003,29(13);31-33
[39]何强,何英.MATLAB扩展编程.北京;清华大学出版社,2002
[40]金晶.轮椅机器人语音识别、控制技术的研究;[硕士学位论文].苏州;苏州大学,2006
[41]魏力.嵌入式语音识别系统的研究;[硕士学位论文].武汉;武汉理工大学,2006
[42]崔健.语音识别技术的研究及应用;[硕士学位论文].沈阳;东北大学,2006
[43]蔡莲红,黄德智.现代语音技术基础与应用.北京;清华大学出版社,2003
[44]张有为,甘俊英,何强等.人机自然交互.北京;清华大学出版社,2003
[45]罗亚非.凌阳16位单片机应用基础.北京;国防工业出版社,2004
[46]宋超,姜力,赵大威.基于SPCE061A的多自由度假手语音控制系统的研究.机械与电 子,2006(10);40-42
[47]张培仁,张志坚,高修峰.十六位单片机微处理器原理及应用(凌阳SPCE061A).北京;清华大学出版社,2005
[48]颜利彪,范蟠果.基于单片机的简易智能电动车.电子技术,2004(4);8-10
[49]叶聪红,徐文龙,王效杰等.两轮驱动小车系统的设计实现及其路径规划.航天制造技术,2005(5);15-21
[50]王晓明.电动机的单片机控制.北京;北京航空航天大学出版社,2002
[51]侯媛彬,袁益民,崔汉平.凌阳单片机原理及其毕业设计精选.北京;科学出版社,2006
[52]凌阳科技教育推广中心.SPCE061A在语音识别机器人中的应用.http;//www.unsp.com/app/html/2006424100102.shtml,2006-4-24