摘要
背景近几年来,利用不同思维作业脑电信号作为实现脑-机接口(BCI)的一种新型人机接口技术,已受到了广泛的关注。BCI是在人脑和计算机或其它电子设备之间建立的一种直接的信息交流和控制通道,它是不依赖于脑的正常输出通路(外周神经系统及肌肉组织)的全新的信息交流系统。它的一个最重要用途是为那些思维正常但有严重运动障碍的患者提供语言交流和环境控制途径,以提高其生存质量;另外,BCI技术在特殊作业和军事环境等领域也有潜在的应用价值,还可以作为一种新的信息交流控制手段和娱乐方式。BCI正成为脑科学、康复工程、生物医学工程及人机自动控制研究领域的一个研究热点。目前,BCI的研究正处于探索阶段,现有BCI系统的识别正确率和识别速率、性能稳定性等关键技术还需要解决。
BCI系统本质上就是识别特定的脑电模式,然后按照预设规则将获取的脑电转化为外部信息。实际应用中,实验设计、信号获取、算法选择(特征提取、特征分类)等因素都会影响最后的结果。因此,BCI系统全面建立过程中,实验设计和信号获取是非常重要的第一步,也是获得良好结果的前提。目前,获取脑电的方式主要是非侵入式方法,它有3种类型:①在思维作业期间,不进行任何刺激情况下,通过头皮电极记录脑电;②利用单次视觉诱发脑电获得脑电;③在感觉运动皮层区,利用μ节律的同步化和去同步化获得脑电。本文主要采用第一种方式获取BCI的控制信号,即通过想象左右手运动获取思维脑电。
本文开展了基于想象左右手运动思维脑电信号的BCI系统控制信号采集、特征提取与识别分类研究。本研究目的是探索一种实用的基于想象运动思维脑电的BCI方式,通过寻找合适的信号处理方法,来提取最能反映不同思维的脑电特征,以得到准确的BCI系统外部控制命令,从而提高BCI系统通讯的识别正确率,为最终实现BCI应用奠定比较坚实的理论和实验基础。
方法基于人在不同思维时表现的基本特性,本文阐述了基于想象左右手运动思维脑电信号用于BCI的原理。利用双计算机和Active One(生理信号测量系统)建立了一个基于想象左右手运动思维脑电的BCI信号采集实验系统。本研究采用头皮电极记录大脑皮层的自发脑电,该记录方式无创,使用者无需训练。通过软件编程的方式产生实验提示控制模式,屏幕上随机地出现向左或向右的箭头,受试者作出相应的选择,从而进行按键。对6名健康受试者进行3种不同时段(箭头出现2s、1s和0s后提示按键)情况下想象左右手运动思维作业的信号采集实验。随后,对实验数据进行离线处理和分析:通过对几种思维脑电信号处理方法的比较研究,提出利用小波多分辨率结合统计特性分析方法提取反映不同思维作业的脑电特征,同时研究了几种可用于本研究信号提取的基本小波函数;在对信号的模式识别方法进行比较深入而系统地研究基础上,提出了利用前向反馈神经网络(BP神经网络)方法识别分类这些特征。
结果对所有受试者三种情况下平均延缓时间Δt_2、Δt_1和Δt_0分析发现,Δt_0与Δt_1和Δt_2之间均存在显著性差别(p<0.05),而Δt_1与Δt_2之间没有显著差别(p>0.05);三种情况下(特别在t=1s时),实际按键前大概0.5~1s左右,想象左右手运动的思维脑电特征信号发生了明显改变,且这些特征存在明显不同。
本研究得到的识别分类结果:在三种不同时段情况下,当训练集和测试集为同一受试者时(即同一受试者的部分样本作为训练集,其他部分样本作为测试集),比如待识别分类的特征向量集为C3通道,三种情况下测试集的平均分类正确率分别达到65.00%、86.67%和72.00%,最大为90.00%,这些结果表明对于同一受试者而言可以获得比较理想的分类效果。不过,训练集和测试集为不同受试者时(即某个受试者作为训练样本集,其他受试者为测试样本集),对其他受试者测试的识别分类正确率的平均值仅为60.00%左右,该结果表明:不同个体之间存在一些差异。
在同一时段情况下(如箭头出现t=1s后提示按键),当特征向量集分别选择C3、C4、CZ和这三个电极合并时,其识别分类正确率分别为86.67%、76.67%、70.00%和65.00%。该结果表明,特征向量选择单独某个电极比这些电极合并时得到的分类率要高些。结论本文设计三种情况下获取BCI信号的实验方案是可行的;本研究提出的在小波变换域中提取特征的方法可以有效地实现信号的去噪、降维和特征提取,是一种十分有效的特征提取方法;采用基于前向反馈神经网络(BP神经网络)的识别分类方法来识别想象左右手运动思维脑电特征,可以得到比较理想分类正确率和BCI系统外部控制信号,是一种有效的且易实现的信号识别方法。
在选择合适通道构建特征向量集基础上,箭头出现1s后提示按键(随机按键)这种情况,可以获得更高的识别分类正确率。这说明通过合理实验设计获取的信号有助于识别正确率的提高;本研究提出的特征作为BCI系统外部装置控制信号是可行的,为BCI系统中不同思维任务的特征提取与识别分类提供了一种新的思路和方法。
Background Recent years, a novel kind of human-computer interface technology using EEG signal of different mental tasks as to achieve Brain-computer interface (BCI) has been explored widely. BCI is a direct and fire-new system of information exchange and control channels between brain and computer or other electronic equipment, which do not depend on the brain’s normal output channels (peripheral nerves and muscles). One of the most important purposes for the BCI comes mainly from the hope that the system can provide language communication and environmental control for those with severe motor disabilities but normal thoughts, and improve the quality of their lives. What’s more, the BCI technology has possible applications in other fields, such as special workover and military affairs, and it will be a new way for exchange or control of information and entertainment. BCI research has drawn attention of scientists in brain-science research, rehabilitation engineering, and biomedical engineering or human machine automatic control. Currently, the BCI technology is still under development. The key point technique need to be solved for the current BCI,such as correct rate and speed of recognition, stability of performance, et al.
BCI essentially recognize the specific pattern of EEG, and translate the EEG into the external information according to the regulation in advance. In practical application, it is very important that we decide to carry out the experimental design, the acquisition of signal (collection of datum) and the arithmetic selection (feature extraction and classification of EEG signal), which also affect the final results. Therefore, the design of experiment and the acquisition of signal are the first step in the establishment of entire BCI and the premise of getting the good result. At present, non-invasive way is the primary way for the acquisition of EEG, which have three types:①EEG was recorded by the scalp electrode under the condition of no stimulation during the mental tasks;②EEG was gotten by the single visual evoked potential stimulus;③EEG was obtained by Event-Related Desynchronizations and Event-Related Synchronization ofμrhythm in the region of sensorimotor cortex. The first type of recording method was taken in our study to acquire the control signal for BCI’s construction, that is, EEG of mental tasks was obtained under the thoughts activity of imaging left-right hands movement.
In present research, collection, feature extraction and recognization classification of control signal based on EEG under the imaging left-right hands movement have been carried out for the study on BCI system construction. The purpose of our study is to explore a practical way of BCI based on mental EEG by imagining movement, find a suitable method of signal processing to extract different EEG feature, which can get precise external control command of BCI system, accordingly, to enhance the correct rate of recognition for BCI system, all of which establish a substantial theory and experiment foundation for the final application of BCI.
Methods The principle of BCI construction based on the mental EEG under the thoughts activity of imaging left-right hands movement was explained according to the basic features in the different thoughts. The experimental system of signal collection for BCI based on mental EEG of imaging left-right hand movement was established by the utilizing double-computer and Active One (biopotential measurement system).In our study, scalp electrode was adopted to record EEG in cerebral cortex. This recording skill is no traumatic and the users don’t need to be trained. The leftward or rightward arrow occurred randomly on screen according to software programming produced the command mode of experimental cue. Then, the subjects gave a homologous selection so that they can press the key. In this paper, the different mental tasks for imaging left-right hands movement from 6 subjects were studied in the experiment of signal collection at three different time section (hint keying after arrow 2s, 1s and 0s). Then, the off-line experimental data were processed and analyzed: By studying several processing methods of mental EEG signal and comparing of each other, the extraction method of EEG features, which can reflect different mental tasks by utilizing the method of combining wavelet multitude resolution analysis with statistics characteristics analysis, was introduced. Meanwhile, several basic wavelet functions that can be used to extract signal were proposed in our study. On the base of the pattern recognition methods for signal deeply compared and systematically studied, these features were recognized and classified by using Feed-forward Back-propagation Neural Network (BP-NN).
Results Average delay timeΔt_2、Δt_1 andΔt_0 for all subjects in three different time section were analyzed, we discovered that there is significant difference (p<0.05) betweenΔt_0 andΔt_2 , betweenΔt_0 andΔt_1, but there is no significant difference (p>0.05) betweenΔt_2 andΔt_1. At three circumstances (at t=1s particularly), there are obviously different features for imaging left-right hands movement about 0.5~1s before practical movement, these features have significant difference.
The main results about recognization and classification were obtained as the fellow: Under the condition of three different time section, when the train sample and test sample come from the same subject (that is, some sample of a subject are the train, the others of the subject are the test), for example, feature vectors to be recognized and classified all come from the C3 passage, the average correct rate of classification for the test sample at three conditions are 65.00%、86.67% and 72.00% respectively, the maximum is 90.00%. These results suggest that ideal classification effect relatively for the same subject were obtained. However, the train sample and the test sample come from different subjects (that is, one subject is the sample of train, the others are the sample of test), the correct rate for others subjects’the test are about 60.00%. These results indicate that there is still some difference among the subjects.
Under the condition of the same time section (as hinting keying after arrow 1s), the C3、C4、CZ and the incorporation of three electrodes were selected as feature vectors, the average correct rate of classification for the test sample are 86.67%,76.67%,70.00%and 65.00% respectively. These results showed that the classification rate obtained from feature vectors selecting certain an electrode was high more than the incorporation of these electrodes.
Conclusion In this paper, it was feasible to design experimental project for acquiring signals of BCI at three circumstances. The extractive method of feature vectors in the domain of wavelet transform, which can effectively remove noise, decrease dimension and extract feature of signal,was proposed and confirmed as a effective and practical technique. The ideal correct rate of classification relatively and the external control signal of BCI system were obtained through using the pattern recognition classification method based Feed-forward Back-propagation Neural Network (BP-NN) to recognize features of mental EEG for imaging left-right hands movement. It is an available method of signal pattern recognition, which is come true easily.
On the base of selecting suitable passage to construct feature vectors, we got higher correct rate of recognization and classification under hinting keying after arrow about 1s. This shows it was helpful to increase the correct rate by reasonable experimental design. The feature extracting method proposed in this study has been proven feasible to be used as external control signals for BCI system. This study has provided new ideas and methods for feature extraction and classification of different mental tasks for BCI.
引文
[1]Wolpaw JR, Birbaumer N, McFarland J, et al. Brain-computer interface for communication and control[J]. Clin Neurophysiol, 2002; 113(6):767-791.
[2]高上凯, 张志广, 高小榕, 洪波, 杨福生. 基于脑-机接口技术的新型医疗器械. 中国 医疗器械杂志, 2006; 30(2):79-82.
[3]伍亚舟, 吴宝明, 何庆华. 基于脑电的脑-机接口系统研究现状. 中国临床康复, 2006; 10(1):147-150.
[4]Wolpaw JR, Birbaumer N, Heetderks WJ, et al. Brain-computer interface technology: a review of the first international meeting[J]. IEEE Trans Rehab Eng, 2000; 8(2):164-173.
[5]何庆华. 基于视觉诱发电位的脑-机接口实验研究. 博士论文, 2003 年.
[6]Vaughan TM, Wolpaw JR, Donchin E. EEG-based communication: prospects and Problems [J]. IEEE Tran Rehab Eng, 1996; 4(4):425-430.
[7]程明, 高上凯, 张琳. 基于脑电信号的脑-计算机接口[J]. 北京生物医学工程, 2000; 19(2):113-118.
[8]Chen YL, Tang FT, Chang WH, et al. The new design of an infrared-controlled human- computer interface for the disabled. IEEE Trans Rehab Eng, 1999; 7(4):474-481.
[9]Smith BT, Mulcahey MJ, Betrz RR. Development of an upper extremity FES system for individuals with C4 tetraplegia[J]. IEEE Trans Rehab Eng, 1996; 4(4):264-270.
[10]Ferguson KA, Polando G, Kobetic R, et al. Walking with a hybrid orthosis system[J]. Spinal Cord, 1999; 37(11):800–804.
[11]Faisal Karmali, Mark Polak, Aleksandar Kostov.Environmental control by a Brain- computer interface. Proceedings of the 22nd Annual EMBS International Conference. Chicago IL EMBS, 2000; 4:2990-2992.
[12]Ikeda A, Shibbasaki H. Invasive recording of movement-related cortical potentials in humans[J]. J Clin Neurophysiol, 1992; 9(4):409-520.
[13]SP Levine, JE Huggins, SL Bement, et al. Identification of EEG patterns as a basis for a direct brain interface[J]. Clin Neurophysiol, 1999; 6(5):439-447.
[14]SP Levine, JE Huggins, SL BeMent, et al. A direct brain interface based on event-related potentials[J]. IEEE Trans Rehab Eng, 2000; 8(2):180-185.
[15]Ernane JX. Costa, Eurvaldo F. Cabral Jr. EEG-based discrimination between imagination of left and right hand movements using adaptive gaussian representation[J]. MedicalEngineering & Physics, 2000; 22(5):345-348.
[16]Pregenzer M, Pfurtsecheller G, Flotzing D. Automated feature selection with a distinction sensitive vector quantizer[J]. Neurocomputing, 1996; 11(1):19-29.
[17]Guger C, Ramoser H, Pfurtsecheller G. Real-time EEG Analysis with subject-specific spatial patterns for brain-computer interface (BCI) [J]. IEEE Trans Rehab Eng, 2000; 8(4):447-456.
[18]杨坤德, 田梦君, 张海南等. 脑-计算机接口技术的研究进展[J]. 生物医学工程杂, 2004; 21(6):1024-1027.
[19]Vaughan TM, Heetderks WJ, Trejo IJ, et a1. Brain-computer interface technology: A review of the second international meeting[J]. IEEE Trans Rehab Eng, 2003; 11(2):94-107.
[20]Dennis J, McFarland, William A, Sarnacki, et al. Brain-computer interface (BCI) operation: optimizing information transfer rates. Biological Psychology, 2003; 63(3):237-251.
[21]Wang Yijun, Wang Ruiping, Gao Xiaorong, et.al. Brain-computer Interface based on the High-frequency Steady-state Visual Evoked Potential. 2005 First International Conference on Neural Interface and Control Proceeding: 37-39.
[22]Schalk G, McFarland DJ, Hinterberger T, et al. BCI2000: A general purpose BC1 system, http://www.bci2000.org.
[23]何庆华, 彭承琳, 吴宝明. 脑-机接口技术研究方法[J]. 重庆大学学报, 2002; 25(12): 106-109.
[24]Dornhege G, blankertz B, Gurio G. Speeding up classification of multi-channel brain- computer spatial patterns for slow cortical potentials [A]. In: Proceeding of the 1st intertional IEEE EMBS Conference on Neural Engineering[C], 2003; 594-597.
[25]Donchin E, Spencer KM, ijesinghe R. The mental prosthesis: Assessing the speed of a P300-based brain–computer interface[J]. IEEE Trans Rehab Eng, 2000; 8(2):174-179.
[26]Sutter Erich E. The brain response interface: communication through visually-induced electrical brain responses[J]. Journal of Microcomputer Applications, 1992; 15(2):31-45.
[27]Pfurtscheller G, Neuper C, Guger C, et al. Current trends in Graz brain–computer interface (BCI) research[J]. IEEE Trans.Rehab.Eng, 2000; 8(2):216-219.
[28]Birbaumer N, Elbert T, Canavan A, et al. Slow potentials of the cerebral cortex and behavior[J]. Physiological Reviews, 1990; 70(1):1-41.
[29]Keirn ZA, Aunon JI. A new mode of communication between man and his surroundings [J]. IEEE Trans Biomed Eng, 1990; 37(12):1209-1214.
[30]Blankertz B, BCI competition 2003. http://ida.first.gmd.de/blanker/competition.
[31]2005 First International Conference on Neural Interface and Control Proceeding, 26-28 May 2005, Wuhan, China.
[32]Birbaumer N, Hinterberger T, Kubler A, et al. The Thought-Translation Device (TTD): neurobehavioral mechanisms and clinical outcome[J]. IEEE Trans Rehab Eng, 2003; 11(2):120-126.
[33]Sterman MB. Basic concepts and clinical findings in the treatment of seizure disorders with EEG operant conditioning. Clin Neurophysiol, 2000; 31(1):45–55.
[34]杨福生, 高上凯. 生物医学信号处理. 北京:高等教育出版社, 1989 年.
[35]Pfurtscheller G, Muller GR, Pfurtscheller J. ‘Thought’-control of functional electrical stimulation to restore hand grasp in a patient with tetraplegia[J]. Neuroscience letters, 2003; 351(1):33-36.
[36]Pfurtscheller G, Pichler-Zalaudek K, Neuper C. Handbook of electroenceph and clin neurophysiol[J]. Amsterdam The Netherlands: Elsevier, 1999; 6:245-268.
[37]Jeannerod MJ. Mental imagery in the motor context[J]. Neuropsy-chologia. 1995; 33 (11):1419-1432.
[38]诸强, 王明时. 左右手运动想象脑电模式的识别[J]. 生物医学工程学杂志, 2004; 21(6):1031-1034.
[39]李坤, 周晓兰, 唐希雯, 张道信, 吴小培. 感知器算法在运动想象脑电模式识别中的应用[J]. 计算机工程与应用, 2006; 42(25):230-232.
[40]Blankertz B, Müller KR, Curio G, et al. The BCI competition 2003: Progress and perspectives in detection and discrimination of EEG single trials[J]. IEEE Trans on Biomedical Eng, 2004; 51(6):1044-1051.
[41]Ramoser H, Muller, Gerking J, Pfurtscheller G. Optimal spatial filtering of single trial EEG during imagined hand movement[J]. IEEE Trans Rehab Eng, 2000; 8(4):441-446.
[42]陈强, 彭虎, 江朝辉等. 独立式脑-计算机接口信号处理和识别方法研究[J]. 北京生物医学工程, 2005; 24(4):254-257.
[43]F Sharbrough, GE Chatrian, RP Lesser, et al. American electroencephalographic society guidelines for standard electrode position nomenclature[J]. J Clin Neurophysiol, 1991; 8(2):200–202.
[44]Ramoser H, Muller, Gerking J, Pfurtscheller G. Optimal spatial filtering of single trial EEG during imagined hand movement[J]. IEEE Trans Rehab Eng, 2000; 8(4): 441-446.
[45]吴晓彬, 邱天爽. 基于时频分析的EEG信号分析处理方法研究进展[J]. 国外医学生物医学工程分册, 2004; 27(6):321-326.
[46]王兆源, 周龙旗.脑电信号的分析方法[J]. 第一军医大学学报, 2000; 20(2):l89-190.
[47]Pfurtscheller G, N euper C. Motor imagery and direct brain-computer communication [J]. Proceedings of the IEEE, 2001; 89(7):1123-1134.
[48]陈香, 杨基海, 叶石朱等. 高频成分在思维作业脑电信号分类中的应用. 生物医学工程学杂志. 2005; 22(6):1259-1263.
[49]McFarland DJ, Lynn M, McCane, et al. Spatial filter selection for EEG-based communication[J]. Electroencephalography and Clinical Neurophysiology, 1997; 103(3):386-394.
[50]DO Walter. A posteriori Wiener filtering of average evoked response, Electroenceph[J]. Clin Neurophysiol, 1969; (3)27:61-70.
[51]DJ Doyle. Some comments on the use of Wiener filtering for the estimation of evoked potentials. Electroenceph Clin Neurophysiol, 1975; 38(5):533-534.
[52]De Weerd JPC. A posteriori time-varying filtering of averaged evoked potentials. Introduction and conceptual basis, Biol Cybernet, 1981; 41:211-222.
[53]Yu KB, McGillem CD. Optimum filter for estimating evoked potential waveforms. IEEE Trans Biomed Eng, 1983; 30(11):730-737.
[54]B.维德罗, SD.史蒂恩斯 著, 王永德, 龙宪惠 译. 自适应信号处理. 成都:四川大学出版社. 1989 年第 1 版.
[55]Thakor NV. Adaptive filtering of evoked potentials. IEEE Trans Biomed Eng, 1987; 34(1): 6-11.
[56]Vaz CA, Thakor NV. Adaptive Fourier estimation of time-varying evoked potentials. IEEE Trans Biomed Eng, 1989; 36(4): 448-455.
[57]Thakor NV, Vaz CA, McPherson RW, Hanley DF. Adaptive Fourier series modeling of time-varying evoked potentials: study of human somatosensory evoked response to etomidate anesthetic. J Electroenceph Clin Neurophys, 1991; 80(2):108-118.
[58]Laguna P, Jane R, Meste O, et.al. Adaptive filter for event-related bioelectric signals using an impulse correlated reference input: comparison with signal averaging techniques. IEEE Trans Biomed Eng, 1992; 39(10): 1032-1044.
[59]Nishida S, Nakamura M, Shibasaki H. Method for single-trial recording ofsomatosensory evoked potentials. J Biomed Eng, 1993; 15(5): 257-262.
[60]何宏彬, 赵富强, 王齐琳. 用于思维脑电提取的自适应噪声抵消算法. 北京生物医学工程, 1997; 16(3):161-166.
[61]Bell AJ, Sejnowske TJ. An information maximization approach to blind separation and blind deconvolution. Neural Computation, 1995; 7(6): 1129-1159.
[62]Amari S, Cichocki A, Yang H. A new learning algorithm for blind signal separation, Advances in Neural Information Processing Systems, 1996; 8:757-763.
[63]Cardoso JF, Laheld BH. Equivariant Adaptive Source Separation. IEEE Transactions on Signal Processing, 1996; 44(12):3017-3030.
[64]Jung TP, Humphries C, Lee TW, et al. Removing electroencephalographic artifacts: comparison between ICA and PCA. Neural Networks for Signal Processing VIII,31 Aug-2 Sep,1998: 63-72.
[65]张韫, 侯晓萌, 张旭. 基于独立分量和功率谱分析的VEP提取. 医疗设备信息, 2004; 19(3):10-13.
[66]杨福生, 洪波, 唐庆玉. 独立分量分析及其在生物医学工程中的应用[J]. 国外医学 生物医学工程分册, 2000; 23(3): 129~134.
[67]郭双, 王汝霖, 王怀阳. 独立分量分析在脑电信号识别方面的应用[J]. 山东生物医 学工程, 2002; 21(4):1-4.
[68]Aapo Hyavrinen, Erkki Oja. Independent Component Analysis: A Tuto rial. http://www.cis.hut.fi/projects/ica/1999
[69]郭晓静, 吴小培, 张道信等. 基于独立分量分析的脑电消噪与特征提取系统仿真学 报, 2003; 15(2):287-289.
[70]龙飞, 吴小培, 范羚. 基于独立分量分析的脑电噪声消除. 生物医学工程学杂志, 2003; 20(3):479-483.
[71]Nai-Jen, Huan, Ramaswamy Palaniappan. Classification of mental tasks using fixed and adaptive autoregressive models of EEG [J]. Proceedings of the 26th Annual International Conference of the IEEE EMBS San Francisco, CA, USA. September1-5, 2004.
[72]Anderson Cw, Stolz Ea, Shamsunder S. Multivariate autoregressive models for classification of spontaneous electroencephalogram during mental tasks[J]. IEEE Trans on Biom Eng, 1998; 45(3): 277-286.
[73]Schloegl A. Using adaptive autoregressive parameters for a brain computer interfaceexperiment. Proceedings of IEEE EMBS Conference[J], 1997; 4 (11): 1533-1535.
[74]胡昌华, 张军波, 夏军, 张伟. 基于 MATLAB 的系统分析与设计-小波分析. 西安电子科技大学出版社, 1999.
[75]郑治真, 沈萍, 杨选辉, 万玉莉. 小波变换及其 MATLAB 工具的应用[J]. 地震出版社, 2001.
[76]Mallat SG. A theory for multiresolution signal decomposition: the wavelet representation [J]. IEEE Trans PAMI, 1989; 11(7):674-693.
[77]Mallat SG. Multiresolution approximations and wavelet orthonormal bases of L2(R)[J]. Trans of the American Mathematical Society, 1989; 313(1):68-87.
[78]Daubechies I. The wavelet transform: time-frequency localization and signal analysis[J]. IEEE Trans on Information Theory, 1990; 36(5):961-1005.
[79]Daubechies I. Ten lectures on wavelet SIAM montpelier. Vermont Capital City Press, 1992; 167-249.
[80]杨帮华, 严国正 ,鄢波. 基于离散小波变换提取脑-机接口中脑电特征[J]. 中国生物医学工程学报, 2006; 25(5):518-522.
[81]Andrew C, Pfurtscheller G. Event-related coherence as a tool for studying dynamic interaction of brain regions[J]. Electroencephalography and clinical Neurophysiology, 1996; 98(2):144-148.
[82]裴晓梅, 和卫星, 郑崇勋. 基于脑电复杂度的意识任务的特征提取与分类[J]. 中国生物医学工程学报, 2005; 24(4):421-425.
[83]易东, 王文昌, 尹全焕等. 军事医学统计学. 西安: 第四军医大学出版社, 2006.
[84]孙即祥, 王晓华, 钟山等. 模式识别中的特征提取与计算机视觉不变量[M]. 北京:国防工业出版社, 2001.
[85]赵喜林, 赵喜玲, 江祥奎. 模式识别方法及其比较分析. 信阳农业高等专科学校学 报, 2004; 14(3):37-40.
[86]杜树新, 吴铁军. 模式识别中的支持向量机方法[J]. 浙江大学学报(工学版), 2003; 37(5):521-527.
[87]赵晖, 荣莉莉, 李晓. 一种设计层次支持向量机多类分类器的新方法[J]. 计算机应用研究, 2003; (5):34-37.
[88]Apnik V. The nature of statistical learning theory[M]. New York: Springer, 1995.
[89]Bahlmann C, Hassdonk B, Burkhardt H. On-line Handwriting Recognition with Support Vector Machines: A Kernel Approach[C]. Ontario: Proc. of the 8th Int.Workshop on Frontiers in Handwriting Recognition, 2002; 49-54.
[90]Jonsson K, Kittler J, Matas YP.Support Vector Machines for Face Authentication[J]. Journal of Image and Vision Computing, 2002; 20(2): 369-375.
[91]Joachims T. Text Categorization with Support Vector Machines: Learning with Many Relevant Features[C]. Proc of the 10th European Conf.Machine Learning, 1999; 137-142.
[92]Ma C, Randolph MA, Drish J. A Support Vector Machines-based Rejection Technique for Speech Recognition[C]. Proceedings of IEEE Int. Conference on Acoustics, Speech, and Signal Processing, 2001; 381-384.
[93]邵才瑞, 李洪奇, 张福明. 句法模式识别及其在地层对比中的应用[J]. 中国海上油气(地质), 2000; 14(5):363-366.
[94]郑南宁. 计算机视觉与模式识别[M]. 北京: 国防工业出版社, 1998.
[95]沈清, 汤霖. 模式识别导论[M]. 长沙: 国防科技大学出版社, 1991.
[96]刘建成, 蔡湛宇. 基于模糊神经网络的肌电信号的分析[J]. 中国医疗器械杂志. 1999; 2 3(2): 80-82.
[97]葛毅, 曹宇, 宫琴等. 基于聚类分析和模糊数学的胎儿心电检测[J]. 清华大学学报(自然科学版). 1999; 39 (5): 20-23.
[98]黄国言, 邹沐昌. 模糊模式识别算法在动态心电图 QRS 模板统计中的应用[J]. 燕山大学学报. 2001; 25(3): 263-265.
[99]王建, 李媛, 康景利等. 小波变换和模糊识别技术在人体血压信号处理中的应用[J]. 信息与控制, 2002; 31(1): 93-96.
[100]S Haykin. Neural networks, a comprehensive foundation Macmillan Publishing Company, 113 Sylvan Avenue, Englewood Cliffs, NJ07632, 1994.
[101]李金宗. 模式识别导论[M]. 北京: 高等教育出版社, 1994.
[102]叶世伟, 史忠植(译). 神经网络原理. 北京: 机械工业出版社, 2004.
[103]何振亚. 多维数字信号处理. 北京: 国防工业出版社, 1995.
[104]JM Jain, K Mohiuddin. Artificial neural networks: a tutorial. IEEE Computer, 1996; 3:31-44.
[105]MH Hassoun. Fundamentals of artificial neural networks. MIT, 1995.
[106]T Back, U Hammel, H Schwefel. Evolutionary computation: C omments on the history and current state [J]. IEEE Transactions on Evolutionary Co mputation, 1997; 1(1): 3-17.
[107]赵元淮, 杜晓东. 神经网络在动态心电图中应用[J]. 临床心电学杂志, 2001;10(1):62-62.
[108]刘书朋, 陈俊强. 利用人工神经网络自动判断脑电图中有无癫痫样放电[J]. 中国生物医学工程学报, 2001; 20(2):116-120.
[109]李鲁平, 程敬之. 神经网络方法在脑诱发电位检测中的应用[J]. 国外医学生物医学工程分册, 1996; 19(5):263-267.
[110]金杰, 赵喆, 郝燕. BP神经网络在药物制剂处方优化中的应用[J]. 中国药剂学杂志, 2006; 4(5):219-224.
[111]周开利, 康耀红 编著. 《神经网络模型及其 MATLAB 仿真程序设计》. 北京: 清华大学出版社, 2005,7.
[112]Mirchandani G, Wei C. On hidden nodes for neural nets[J]. IEEE Trans on Circuits and Systems, 1992; 36(5):661-664.
[113]高大启. 有导师的线性基本函数前向三层神经网络结构研究[J]. 计算机学报, 1998; 21(1):80-86.
[114]Pfurtscheller G, Kalcher J, Neuper C, et al. On-Line Eeg classification during externally-paced hand movements using a neural network-based classifier[J]. Electroencephalogr Clin Neurophysiol, 1996; 99(5): 416-425.
[115]许东, 吴铮. 基于 MATLAB6.x 的系统分析与设计-神经网络[M]. 第 2 版, 西安:电子科技大学出版社, 2002; 45-60.
[116]Klimesch WG. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis[J]. Brain Research Reviews, 1999; (29):169-195.
[117]Ramoser H, Muller, Gerking J, Pfurtscheller G. Optimal spatial filtering of single trial EEG during imagined hand movement[J]. IEEE Trans Rehab Eng, 2000; 8(4):441-446.