基于无线传感器网络的声源目标搜寻多机器人系统
|
摘要
声源搜寻机器人是由麦克风阵列、磁阻传感器、无线传感器网络节点、移动小车组成的系统,利用麦克风传感器阵列采集语音信息,利用磁阻传感器测量航向角,能够实现声源识别、声源定位及声源搜寻,最终寻找到目标声源。机器人系统将无线传感器网络、微处理器技术、信号处理和计算机技术融合在一起,在机器人语音控制、危险场合人员搜救等领域有着广阔的应用前景。
系统由3个移动机器人小车组成,3个机器人作为网络节点与上位机(Sink节点)共同组成无线传感器网络。上位机对3个机器人小车的声源识别结果进行融合,从而实现机器人协同工作,在含噪声环境下寻找各自的声源目标。在搜寻过程中,机器人将关键信息以文本格式存储在SD卡中,上位机实时显示各个机器人的航向角和声源方向角等信息,并完成机器人的控制功能。
本文主要从机器人系统的算法理论、硬件电路和软件代码三方面进行了研究,并通过实验验证方案的有效性。
算法理论采用了Mel频率倒谱系数、改进广义互相关和球形插值定位法,分别实现了声源识别、麦克风阵列时延估计和声源定位。此外,依据声源搜寻方法指导机器人完成声源搜寻。
硬件电路基于DSP TMS320F28335平台,功能模块包括外扩RAM存储器,电机驱动控制,麦克风阵列及磁阻传感器信号调理,实时时钟,SD卡存储器,无线传感器网络节点,串口通信。
软件代码由DSP数据处理和控制程序、上位机显示和控制程序及无线传感器网络节点收发程序三部分组成。DSP数据处理和控制程序为系统的核心,完成了算法运行和机器人控制。上位机软件接收和显示机器人发送的航向角、声源方向角、声源识别结果等信息,完成声源识别结果的融合,并实现三个机器人的控制功能。无线传感器网络节点收发程序完成机器人之间及与Sink节点的无线通信,并实现节点(机器人)间距离的测量。
最后对整个声源搜寻系统进行了实验。利用三个音箱分时播放三段含有噪声音乐,三个机器人最终寻找到各自的目标声源,实验结果达到预期设计目标,从而验证了本系统设计的有效性。
Sound source search robot is composed of microphone array sensors, magneto resistive sensor, wireless sensor networks node and mobile robot vehicle. Using microphone array sensors for voice acquisition and magneto resistive sensor for course measurement, robot can accomplish sound source recognition, source localization and source search, and ultimately find sound source target. The robot system combines wireless sensor networks, microprocessor technology, signal processing and computer technology together, and will have a broad application prospects at voice control of robot and person search and rescue in dangerous situations.
The system consists of three mobile robots, which as network nodes and PC (Sink node) compose the wireless sensor network. PC fuses the sound source recognition results of the three robots, can realizes that the three robots working together and searching each sound target in environment with noise. In the process of sound search, robot saves the key message as text file in the SD card memory, PC displays the course, angle of sound and so on information of each robot, and also accomplishes control of the robots.
In this paper, theory of algorithms, hardware and software in three areas of robot system are studied, and the effectiveness of the design program is verified through experiments.
Theory of algorithms adopt Mel-frequency Cepstral coefficients, generalized cross correlation and spherical interpolation positioning method, which are used to achieve sound source recognition, time delay of estimation and source localization respectively. In addition, the sound search method leads robot to complete the task of sound source search.
Hardware platform is based on DSP TMS320F28335, and function modules contain external extended RAM, motor control and drive, signal conditioning of microphone array and the magneto resistive sensors, real-time clock, SD card memory, wireless sensor networks node, serial communication.
Software platform is composed of three parts programs: DSP data processing and control, PC display and control and wireless sensor networks (WSN) node transceiver. Program of DSP data processing and control, which is the core of the system, can accomplish the operation of algorithms and control of robot; Program of PC receives and displays the course, sound source direction angle, results of sound source recognition and so on information which is sent by robots, fuses sound source recognition results of the three robots, and achieves the control of the three robots; Program of WSN node transceiver fulfills wireless communication between the robots and Sink node, and also achieves measurement of the distance between nodes (robots).
Finally, experiments of sound source search system are performed. When the three songs containing noise are played asynchronously through three speakers respectively, the three robots can find their own sound sources eventually, and achieves the desired result and validates the effectiveness of this system.
系统由3个移动机器人小车组成,3个机器人作为网络节点与上位机(Sink节点)共同组成无线传感器网络。上位机对3个机器人小车的声源识别结果进行融合,从而实现机器人协同工作,在含噪声环境下寻找各自的声源目标。在搜寻过程中,机器人将关键信息以文本格式存储在SD卡中,上位机实时显示各个机器人的航向角和声源方向角等信息,并完成机器人的控制功能。
本文主要从机器人系统的算法理论、硬件电路和软件代码三方面进行了研究,并通过实验验证方案的有效性。
算法理论采用了Mel频率倒谱系数、改进广义互相关和球形插值定位法,分别实现了声源识别、麦克风阵列时延估计和声源定位。此外,依据声源搜寻方法指导机器人完成声源搜寻。
硬件电路基于DSP TMS320F28335平台,功能模块包括外扩RAM存储器,电机驱动控制,麦克风阵列及磁阻传感器信号调理,实时时钟,SD卡存储器,无线传感器网络节点,串口通信。
软件代码由DSP数据处理和控制程序、上位机显示和控制程序及无线传感器网络节点收发程序三部分组成。DSP数据处理和控制程序为系统的核心,完成了算法运行和机器人控制。上位机软件接收和显示机器人发送的航向角、声源方向角、声源识别结果等信息,完成声源识别结果的融合,并实现三个机器人的控制功能。无线传感器网络节点收发程序完成机器人之间及与Sink节点的无线通信,并实现节点(机器人)间距离的测量。
最后对整个声源搜寻系统进行了实验。利用三个音箱分时播放三段含有噪声音乐,三个机器人最终寻找到各自的目标声源,实验结果达到预期设计目标,从而验证了本系统设计的有效性。
Sound source search robot is composed of microphone array sensors, magneto resistive sensor, wireless sensor networks node and mobile robot vehicle. Using microphone array sensors for voice acquisition and magneto resistive sensor for course measurement, robot can accomplish sound source recognition, source localization and source search, and ultimately find sound source target. The robot system combines wireless sensor networks, microprocessor technology, signal processing and computer technology together, and will have a broad application prospects at voice control of robot and person search and rescue in dangerous situations.
The system consists of three mobile robots, which as network nodes and PC (Sink node) compose the wireless sensor network. PC fuses the sound source recognition results of the three robots, can realizes that the three robots working together and searching each sound target in environment with noise. In the process of sound search, robot saves the key message as text file in the SD card memory, PC displays the course, angle of sound and so on information of each robot, and also accomplishes control of the robots.
In this paper, theory of algorithms, hardware and software in three areas of robot system are studied, and the effectiveness of the design program is verified through experiments.
Theory of algorithms adopt Mel-frequency Cepstral coefficients, generalized cross correlation and spherical interpolation positioning method, which are used to achieve sound source recognition, time delay of estimation and source localization respectively. In addition, the sound search method leads robot to complete the task of sound source search.
Hardware platform is based on DSP TMS320F28335, and function modules contain external extended RAM, motor control and drive, signal conditioning of microphone array and the magneto resistive sensors, real-time clock, SD card memory, wireless sensor networks node, serial communication.
Software platform is composed of three parts programs: DSP data processing and control, PC display and control and wireless sensor networks (WSN) node transceiver. Program of DSP data processing and control, which is the core of the system, can accomplish the operation of algorithms and control of robot; Program of PC receives and displays the course, sound source direction angle, results of sound source recognition and so on information which is sent by robots, fuses sound source recognition results of the three robots, and achieves the control of the three robots; Program of WSN node transceiver fulfills wireless communication between the robots and Sink node, and also achieves measurement of the distance between nodes (robots).
Finally, experiments of sound source search system are performed. When the three songs containing noise are played asynchronously through three speakers respectively, the three robots can find their own sound sources eventually, and achieves the desired result and validates the effectiveness of this system.
引文
1袁正午,肖旺辉.改进的混合MFCC语音识别算法研究.计算机工程与应用. 2009, 45(33): 108-110
2李建中,李金宝,石胜飞.传感器网络及其数据管理的概念、问题与进展.软件学报. 2003, 14(10): 1717-1727
3任丰原,黄海宁,林闯.无线传感器网络.软件学报. 2003, 14(7): 282-291
4 Noury N, Herve T, Rialle V, Virone G, Mercier E. Monitoring behavior in home using a smart fall sensor. Lyon: IEEE Computer Society, 2000: 607-610
5王换招,范琳,王海,李增智.无线传感器网络多目标跟踪数据融合.西安交通大学学报. 2006, 40(10): 1043-1046
6马瑞恒,盛晓红.无线传感网络中分布式粒子滤波的目标追踪算法.解放军理工大学学报. 2006, 7(10): 421-425
7 0lson H, Belar H. Phonetie Typewriter. Aeoust. Soe.Am, 1956, 28(6): 1072-1081
8 Rabiner L, Juang B. Fundamentals of Speeeh Recognition. Prentice-Hall. 1993
9 T.B.Martin, A.L.Ne1Son, H.J.Zadel. Speeeh recognition by feature abstraction techniques. AirForeeAvionics Lab. 1964
10 Lesser V, Fennell R, Erman L, etal. The Hearsay Speeeh Understanding System. IEEETrans. Aeousties, SPeeeh, SignalProc. 1975, ASSP-23(l): 11-24
11王炳锡,屈丹,彭煊.实用语音识别基础.国防工业出版社, 2005: 56-57
12 Flanagan JL. Bandwidth design for Speech-seeking microphone arrays. IEEE Intemational Conferenee on Aeoustics, Speeeh and Signal Proeessing. 1985, 10: 732-735
13 Brandstein MS, Silverman HF. A Practical methodology for Speeeh source localization with microphone arrays. Computer speeeh and Language. 1997, 11(2): 91-126
14 Bernard Widrow, Fa-Long Luo. Microphone arrays for hearing aids. Speeeh Communication. 2003, 39(l2): 139-146
15邵怀宗,林静然等.基于麦克风阵列的声源定位研究.云南民族大学学报(自然科学版). 2004, 13(4): 255-258
16吴俣.基于麦克风阵列的声源定位技术的研究.硕士论文.电子科技大学. 2003
17王振涛.郝忠孝.贺洪江.基于传声器阵列的声源定位系统的研究.华北电力大学学报, 36(5): 103-105
18崔玮玮,曹志刚等.声源定位中的时延估计技术.数据采集与处理. 2007, 22(1): 90-99
19张贤达.现代信号处理.清华大学出版社, 2002
20韩纪庆等.语音信号处理.清华大学出版社, 2004
21董志峰.汪增福.基于动态MFCC的说话人识别算法.模式识别与人工智能. 2005, 18(5): 596-601
22 Molau S, Pitz M, Schluter R, Ney H. Computing Mel-Frequency Cepstral Coefficients on the Power Spectrum. Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing. 2001: 73-76
23 Murthy H, Bcaufays F, Heek L, etal. Robust Text-Independent Speaker Identification over Telephone Channels. IEEE Trans on Speech and Audio processing. 1999, 7(5): 554-568
24赵力.语音信号处理.机械工业出版社, 2003
25陈勇,屈志毅等.语音特征参数MFCC的提取及应用.湖南农业大学学报(自然科学版). 2009, 35(1): 106-107
26易克初,田斌,付强.语音信号处理.国防工业出版社, 2000: 249-264
27 S. Umesh, L. Cohen, D. Nelson. Frequency Warping and the Mel Scale. IEEE Signal Processing Letters. 2001, Vol.9(3): 104-107
28李弼程等.模式识别原理及应用.西安电子科技大学出版社, 2008: 231-232
29甄斌,吴玺宏,刘志敏,迟惠生.语音识别和说话人识别中各倒谱分量的相对重要性.北京大学学报(自然科学版). 2001, 37(3): 371-378
30杨祥清,汪增福.基于麦克风阵列的三维声源定位算法及其实现.声学技术. 2008, 27(2): 260-265
31王毅,吴长奇,胡双喜. TDOA中几种时延估计算法的比较.工程实践及应用技术. 2008, 34(1): 52-55
32徐胜.基于智能麦克风阵列的说话人跟踪技术.电子科技大学硕士论文. 2009: 3-11
33周峰.室内麦克风阵列声源定位算法研究和实现.复旦大学硕士论文. 2009: 17-22
34杜要峰,尹雪飞等.一种修正的近场声源定位时延估计方法.声学基础. 2010, 34(2): 47-50
35周浩洋.基于麦克风阵列的声源定位方法研究.大连理工大学硕士论文. 2002: 30-39
36张莉,吕明.基于特征分解的一步最小二乘声源定位算法.电声技术. 2007, 31(5): 42-45
37苏奎峰,吕强等. DSP原理及C语音程序开发.北京航空航天大学出版社, 2008: 73-85
38于金涛,韩轲等.基于CC2430的无线传感器网络系统节点设计.哈尔滨商业大学学报(自然科学版). 2010, 26(2): 192-195
39彭启琮. TI DSP集成化开发环境(CCS)使用手册.清华大学出版社, 2007: 21-54
40刘和平等. DSP指令和编程指南.清华大学出版社, 2005
41田黎育,何佩琨. TMS320C6000系列DSP编程工具与指南.清华大学出版社, 2007: 101-121
42杨占昕,邓纶晖等. TMS320C54x系列DSP指令和编程指南.清华大学出版社, 2010
43 David M. Alter. Running an Application from Internal Flash Memory on the TMS320F281x DSP. Texas Instruments Incorporated, 2004: 11~21
44刘瑞新,曹建春等. Visual C++面向对象程序设计教程.机械工业出版社, 2004: 12-42
2李建中,李金宝,石胜飞.传感器网络及其数据管理的概念、问题与进展.软件学报. 2003, 14(10): 1717-1727
3任丰原,黄海宁,林闯.无线传感器网络.软件学报. 2003, 14(7): 282-291
4 Noury N, Herve T, Rialle V, Virone G, Mercier E. Monitoring behavior in home using a smart fall sensor. Lyon: IEEE Computer Society, 2000: 607-610
5王换招,范琳,王海,李增智.无线传感器网络多目标跟踪数据融合.西安交通大学学报. 2006, 40(10): 1043-1046
6马瑞恒,盛晓红.无线传感网络中分布式粒子滤波的目标追踪算法.解放军理工大学学报. 2006, 7(10): 421-425
7 0lson H, Belar H. Phonetie Typewriter. Aeoust. Soe.Am, 1956, 28(6): 1072-1081
8 Rabiner L, Juang B. Fundamentals of Speeeh Recognition. Prentice-Hall. 1993
9 T.B.Martin, A.L.Ne1Son, H.J.Zadel. Speeeh recognition by feature abstraction techniques. AirForeeAvionics Lab. 1964
10 Lesser V, Fennell R, Erman L, etal. The Hearsay Speeeh Understanding System. IEEETrans. Aeousties, SPeeeh, SignalProc. 1975, ASSP-23(l): 11-24
11王炳锡,屈丹,彭煊.实用语音识别基础.国防工业出版社, 2005: 56-57
12 Flanagan JL. Bandwidth design for Speech-seeking microphone arrays. IEEE Intemational Conferenee on Aeoustics, Speeeh and Signal Proeessing. 1985, 10: 732-735
13 Brandstein MS, Silverman HF. A Practical methodology for Speeeh source localization with microphone arrays. Computer speeeh and Language. 1997, 11(2): 91-126
14 Bernard Widrow, Fa-Long Luo. Microphone arrays for hearing aids. Speeeh Communication. 2003, 39(l2): 139-146
15邵怀宗,林静然等.基于麦克风阵列的声源定位研究.云南民族大学学报(自然科学版). 2004, 13(4): 255-258
16吴俣.基于麦克风阵列的声源定位技术的研究.硕士论文.电子科技大学. 2003
17王振涛.郝忠孝.贺洪江.基于传声器阵列的声源定位系统的研究.华北电力大学学报, 36(5): 103-105
18崔玮玮,曹志刚等.声源定位中的时延估计技术.数据采集与处理. 2007, 22(1): 90-99
19张贤达.现代信号处理.清华大学出版社, 2002
20韩纪庆等.语音信号处理.清华大学出版社, 2004
21董志峰.汪增福.基于动态MFCC的说话人识别算法.模式识别与人工智能. 2005, 18(5): 596-601
22 Molau S, Pitz M, Schluter R, Ney H. Computing Mel-Frequency Cepstral Coefficients on the Power Spectrum. Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing. 2001: 73-76
23 Murthy H, Bcaufays F, Heek L, etal. Robust Text-Independent Speaker Identification over Telephone Channels. IEEE Trans on Speech and Audio processing. 1999, 7(5): 554-568
24赵力.语音信号处理.机械工业出版社, 2003
25陈勇,屈志毅等.语音特征参数MFCC的提取及应用.湖南农业大学学报(自然科学版). 2009, 35(1): 106-107
26易克初,田斌,付强.语音信号处理.国防工业出版社, 2000: 249-264
27 S. Umesh, L. Cohen, D. Nelson. Frequency Warping and the Mel Scale. IEEE Signal Processing Letters. 2001, Vol.9(3): 104-107
28李弼程等.模式识别原理及应用.西安电子科技大学出版社, 2008: 231-232
29甄斌,吴玺宏,刘志敏,迟惠生.语音识别和说话人识别中各倒谱分量的相对重要性.北京大学学报(自然科学版). 2001, 37(3): 371-378
30杨祥清,汪增福.基于麦克风阵列的三维声源定位算法及其实现.声学技术. 2008, 27(2): 260-265
31王毅,吴长奇,胡双喜. TDOA中几种时延估计算法的比较.工程实践及应用技术. 2008, 34(1): 52-55
32徐胜.基于智能麦克风阵列的说话人跟踪技术.电子科技大学硕士论文. 2009: 3-11
33周峰.室内麦克风阵列声源定位算法研究和实现.复旦大学硕士论文. 2009: 17-22
34杜要峰,尹雪飞等.一种修正的近场声源定位时延估计方法.声学基础. 2010, 34(2): 47-50
35周浩洋.基于麦克风阵列的声源定位方法研究.大连理工大学硕士论文. 2002: 30-39
36张莉,吕明.基于特征分解的一步最小二乘声源定位算法.电声技术. 2007, 31(5): 42-45
37苏奎峰,吕强等. DSP原理及C语音程序开发.北京航空航天大学出版社, 2008: 73-85
38于金涛,韩轲等.基于CC2430的无线传感器网络系统节点设计.哈尔滨商业大学学报(自然科学版). 2010, 26(2): 192-195
39彭启琮. TI DSP集成化开发环境(CCS)使用手册.清华大学出版社, 2007: 21-54
40刘和平等. DSP指令和编程指南.清华大学出版社, 2005
41田黎育,何佩琨. TMS320C6000系列DSP编程工具与指南.清华大学出版社, 2007: 101-121
42杨占昕,邓纶晖等. TMS320C54x系列DSP指令和编程指南.清华大学出版社, 2010
43 David M. Alter. Running an Application from Internal Flash Memory on the TMS320F281x DSP. Texas Instruments Incorporated, 2004: 11~21
44刘瑞新,曹建春等. Visual C++面向对象程序设计教程.机械工业出版社, 2004: 12-42