1274269 玖、發明說明 【發明所屬之技術領域】 本發明係關於一種人機控制系統及方法,特別指一種 可自動即時分類腦波訊號以控制一受控裝置之腦波訊號分 類方法及腦波訊號驅動之人機控制系統及方法。 5【先前技術】 首先如第一圖所示,人腦中存在許多功能性區域,當 人體發生功能性事件時(如接受光刺激、聽覺刺激、或大 腦指揮肢體運動等)’對應的腦部皮質功能區將有對應之 神經元將被激活而放電,除在鄰近區域造成局部之離子濃 10 度變化,並造成一段距離内之微量電場、磁場變化。所以 上述經由各種發生事件所引發之神經放電過程,可經由諸 如在頭部表面貼附電極量測腦電波、或由腦磁波儀偵測, 因其時間解析度(temporal resolution)高之優點’可福測腦 中所產生各種頻率之訊號以進行腦波的分析。 15 腦波訊號主要分為腦波誘發訊號(averaged evoked response)及腦波律動訊號(brain rhythm)雨大類,分別簡介 如下: 一、腦波誘發訊號: 當人腦接受外界刺激或藉由自身意志而擬執行某一動 20 作時,如第一圖所示,於相對應腦部區域將產生相對應特 定波形之腦波變化,例如執行手部運動時,將有運動相關 腦波訊號 MRP(motor related potential)或 MRF(motor related field)、接受色彩或亮度之視覺刺激時將有 VEP(visual evoked potential)^ VEF(visual evoked field) 5 1274269 而接受聽覺刺激時將有AEP(auditory evoked potential)或 AEF(auditory evoked field)等。若以產生事件之時間點為 基準分析,上述訊號通常與基準點之距離一特定時間間隔 後發生,故稱為時間鎖定(time-locked),且重複測試時, 5 發生的訊號起始相位均大致不變,故稱為相位鎖定(phase-locked) 〇 由此,分析腦波誘發訊號之方法,係多次重複同一事 件(例如閃光或電刺激等事件),並依據事件發生時間作為 基準點,將連續測得之腦波電訊號依每次事件時間為基準 10 ,切割為複數資料段(epoch),再依據各資料段之事件時間 點將各資料段對齊,並經多次平均(一般約100次)以提高 訊雜比。惟由於該習知方法需經多次平均計算,極難達到 即時分析、甚至用以控制周邊裝置(如輪椅)之目的。 另方面,腦波律動訊號雖為時間鎖定但非相位鎖定 15 (time-locked but not phase locked),不適用直接事件平均 之方法,故腦波律動訊號及腦波誘發訊號之分析運算方法 亦有相當差異。 二、腦波律動訊號: 人腦中存在許多功能區域性腦波律動,較為人知者 20 包括:(l)Mu rhythm ··約存在於10〜20Hz頻帶之間,主要 區域為感覺運動區(sensorimotor area),(2)Tau rhythm :約 存在於8〜10 Hz之間,存在區域為上顳葉皮質區(upper temporal lobe),(3)sigma rhythm ··約存在於 7〜9 Hz 之間 ,存在區域為感覺區(sensory area),(4)Alpha rhythm :約 1274269 10Hz,存在區域為枕葉視覺區。上述腦波律動常與特定功 能相關並存在於特定區域,故常被用來作為特定腦功能區 域之功能性分析依據。 如前所述,因腦波律動訊號屬於非相位鎖定 5 (nonphase-locked)的訊號,在事件發生時之起始相位並不 固定,若單純以一般事件觸發電位(event-related potential ,ERP)直接平均,則獲得結果將彼此抵銷。故須採用非 相位鎖定之特殊分析技巧,計算由外界刺激所產生之反應 。目前已有一些腦波律動分析技術相繼發展問世,並已應 10 用於事件觸發腦波律動之變化偵測上。 如 G. Pfurtscheller 等學者於 “Evaluation of event-realted desynchronization (ERD) preceding and following voluntary self-paced movement,” Electroencephalography and clinical neurophysiology,vol: 46,pp: 138-146,1979.及 15 “Patterns of cortical activation during planning of voluntary movement,” Electroencephalography and clinical neruophysiology, vol: 72, pp: 250-258, 1989.兩文獻中,艮p 利用腦電波量測手指按鍵動作(finger tapping task)之事件 觸發非相位鎖定訊號,並使用power method和inter-trial 20 variance方法,計算於事件發生前後所產生之能量變化, 發現受測者運動前,10Hz附近之Mu rhythm有能量遞減 現象,稱為 event_related desynchronization (ERD) 〇 而 GL Pfurtscheller 等學者於“Distinction of different fingers by the frequency of stimulus induced beta 1274269 oscillations in the human EEG/9 Neuroscience letters, vol: 307,pp: 48-52,2001·、“Event-related synchronization (ERS): an electrophysiological correlated of cortical areas at rest,” Electroencephalography and clinical neuro-5 physiology,vol: 83, pp: 62-69,1992·及“Central beta rhythm during sensorimotor activities in man,” Electro-encephalograpy and clinical neurophysiology, vol: 51,pp: 253-264,1981.文獻中,更發現於運動結束後約1秒中, 於20 Hz左右有一能量之反彈(rebound),稱為event-10 related synchronization (ERS) 〇 此外,G. Pfurtscheller亦將ERD與ERS技術應用於 體表電刺激(electric stimulation)所誘發之Mu rhythm變化 上,並發現當接受電刺激後約0.5〜0.7秒,於14〜18Hz左 右有 ERS 發生。R. Hari 利用 temporal spectral evoluation 15 (TSE)方法於腦磁波上計算ERD及ERS,並得到與腦電波 類似的結果。T. Teija利用TSE方法發現於Unverricht-Lundborg type之癲癇病人上執行手部median nerve之電 刺激,會有ERS消失現象。 上述各公開文獻主要皆係利用所執行之事件,均係預 20 先限制所欲分析之事件,並選擇與事件相關之腦部區域表 面佈電極,再偵測事件觸發所引起腦波律動訊號在時間上 之變化情況,作為生理與病理上評估依據。美國第 5,638,826專利案中’更循同一邏輯而進一步利用事件觸 發所引起之腦波律動變化,將手部運動所引起之 9 1274269 erd/ers訊號作為週邊設備之控制訊號。 —然而’上述方法,皆須先選定特定量測通道,進行特 定位置之訊號量測與分析,且預先界定該區域之神經元放 電頻率,將問題簡化至相當狹窄之頻率範圍,故可單純用 5滤'波器渡除與該特定區域的腦波律動不相關之訊號。一則 *此預疋頻寬範15内存有與預設之腦波律動頻率相近之 雜訊’則無法將此種雜訊渡除,故該等方法通常無法於少 數一、兩次試驗内達到極高之訊雜比,而需多次重複試驗 ,再平均處理後才能達到良好之控制訊號品f。另方面, 腦部正常運作過程中所巍含的複雜訊息,並不符合如此簡 化之模型,而此等分析方式勢將浪費諸多有意義的訊息, 無異於買櫝還珠。 由此’右月b針對所擷取之複雜腦波訊號進行分離,搏 除無意義之雜訊,則所有對腦部生理運作進行之研究分析 15將更為順利。更進一步,若能將分離出之成分與預先獲得 之事件資料進行比對,確認某些已知事件的發生,以此作 為人,間之控制界面,將使被操縱之機械裝置更能隨操作 者之意運作,如電影「火狐裡」中所想像之腦波控制飛機 亦將逐漸實現,而受限於神經傳輸障礙(如漸滚人等)之患 20者’亦將可藉由腦波控制周邊物品,提高生活自理能力。 【發明内容】 因此,本發明之-目的,是提供一種可自動解析量得 之腦波訊號成分,並加以分類之方法。 本發明之另-目的,是提供一種可稱除量得之腦波訊 10 1274269 【圖式之主要元件代表符號簡單說明】 1 ••… …·訊號量測裝置 22…… …分解單元 2…·· …·訊號處理裝置 23…… …選擇單元 3 … …·儲存裝置 24…… …重組單元 4 ••… …·受控裝置 25…… …分類單元 5 ····· …·使用者 251… …運算器 11 ••… …·量測電極 252… …分類器 12 …·· …·感測器 41…… …驅動控制器 21 ••… …·七處理單元 42…… …驅動馬達 21BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a human-machine control system and method, and more particularly to a brain wave signal classification method and brain wave signal capable of automatically classifying brain wave signals to control a controlled device. Driven human-machine control system and method. 5 [Prior Art] First, as shown in the first figure, there are many functional areas in the human brain. When the human body has a functional event (such as receiving light stimulation, auditory stimulation, or brain-directed limb movement, etc.), the corresponding brain In the cortical functional area, the corresponding neurons will be activated and discharged, except that in the adjacent area, the local ion concentration is changed by 10 degrees, and a slight electric field and magnetic field change within a certain distance is caused. Therefore, the above-mentioned nerve discharge process caused by various occurrence events can be measured by attaching electrodes to the surface of the head to measure brain waves, or detected by a magnetoencephalograph, because of the advantages of high temporal resolution. The signals of various frequencies generated in the brain are measured for brain wave analysis. 15 Brainwave signals are mainly divided into the averaged evoked response and the brain rhythm (rainin rhythm) rain, which are as follows: 1. Brain wave induced signal: When the human brain receives external stimulation or through its own will When a certain action is to be performed, as shown in the first figure, brain wave changes corresponding to a specific waveform will be generated in the corresponding brain region. For example, when performing hand motion, there will be a motion-related brain wave signal MRP (motor). Related potential) or MRF (motor related field), there will be VEP (visual evoked potential) V VEF (visual evoked field) 5 1274269 when receiving visual stimuli of color or brightness, and there will be AEP (auditory evoked potential) when receiving auditory stimuli or AEF (auditory evoked field) and so on. If the time is based on the time point at which the event is generated, the signal usually occurs after a certain time interval from the reference point, so it is called time-locked, and when the test is repeated, the start phase of the signal occurs. It is roughly unchanged, so it is called phase-locked. Therefore, the method of analyzing brain wave-induced signals repeats the same event multiple times (such as flash or electrical stimulation) and uses the time of occurrence as a reference point. The continuously measured brainwave electrical signal is cut into multiple data segments (epoch) according to the time of each event, and then the data segments are aligned according to the event time points of each data segment, and the average is averaged (generally about 100 times) to improve the signal to noise ratio. However, since the conventional method requires multiple averaging calculations, it is extremely difficult to achieve immediate analysis or even to control peripheral devices such as wheelchairs. On the other hand, although the brain wave motion signal is time-locked but not phase locked, it is not applicable to the method of direct event averaging. Therefore, the analysis methods of brain wave motion signals and brain wave evoked signals are also available. Quite a difference. Second, the brain wave motion signal: There are many functional regional brain wave rhythms in the human brain, more well-known 20 include: (l) Mu rhythm · · about exist between the 10~20Hz band, the main area is the sensory motion zone (sensorimotor Area), (2) Tau rhythm: about 8~10 Hz, the upper temporal lobe is present, and (3) sigma rhythm is about 7~9 Hz. The area of existence is the sensory area, (4) Alpha rhythm: about 1274269 10 Hz, and the area of existence is the occipital vision area. The above-mentioned brain wave motions are often associated with specific functions and exist in specific areas, so they are often used as a functional analysis basis for specific brain functional areas. As mentioned above, since the brain wave motion signal is a nonphase-locked signal, the initial phase is not fixed at the time of the event, if it is simply an event-related potential (ERP). Direct averaging, the results will be offset against each other. Therefore, special analysis techniques such as non-phase locking must be used to calculate the response generated by external stimuli. At present, some brain wave motion analysis techniques have been developed and have been used for event detection of brain wave rhythm changes. Such as G. Pfurtscheller and other scholars in "Evaluation of event-realted desynchronization (ERD) preceding and following voluntary self-paced movement," Electroencephalography and clinical neurophysiology, vol: 46, pp: 138-146, 1979. and 15 "Patterns of cortical Activation during planning of voluntary movement," Electroencephalography and clinical neruophysiology, vol: 72, pp: 250-258, 1989. In both documents, 艮p triggers non-phase locking using events of the finger tapping task. Signal, and using the power method and inter-trial 20 variance method, calculate the energy change produced before and after the event, and find that Mu rhythm near 10Hz has energy degeneration before the subject's motion, called event_related desynchronization (ERD) 〇 And GL Pfurtscheller et al., "Distinction of different fingers by the frequency of stimulus induced beta 1274269 oscillations in the human EEG/9 Neuroscience letters, vol: 307, pp: 48-52, 2001 ·, "Event-related synchronization (E RS): an electrophysiologically correlated of cortical areas at rest," Electroencephalography and clinical neuro-5 physiology, vol: 83, pp: 62-69, 1992· and "Central beta rhythm during sensorimotor activities in man," Electro-encephalograpy and clinical Neurophysiology, vol: 51, pp: 253-264, 1981. In the literature, it was found that in about 1 second after the end of exercise, there is an energy rebound around 20 Hz, called event-10 related synchronization (ERS). In addition, G. Pfurtscheller also applied ERD and ERS technology to Mu rhythm induced by electric stimulation, and found that about 0.5~0.7 seconds after receiving electrical stimulation, there are ERS around 14~18Hz. occur. R. Hari uses the temporal spectral evoluation 15 (TSE) method to calculate ERD and ERS on brain waves and obtain similar results to brain waves. T. Teija used the TSE method to detect electrical stimulation of the median nerve in the Unverricht-Lundborg type epileptic patient, and ERS disappeared. Each of the above publications mainly utilizes the executed events, and both pre-limit the events to be analyzed, and select the surface electrode electrodes of the brain regions associated with the events, and then detect the brain wave motion signals caused by the event triggers. The change in time is used as a basis for physiological and pathological evaluation. In the US Patent No. 5,638,826, the brainwave rhythm changes caused by event triggers are further utilized in the same logic, and the 9 1274269 erd/ers signal caused by the hand movement is used as the control signal of the peripheral device. - However, in the above methods, it is necessary to first select a specific measurement channel, perform signal measurement and analysis at a specific position, and predefine the neuron discharge frequency of the region, simplifying the problem to a relatively narrow frequency range, so it can be used simply 5 filter 'waves to remove signals that are not related to brain wave rhythm in this particular area. A *This pre-frequency band 15 has a noise similar to the preset cranial wave frequency', so this kind of noise can't be removed, so these methods can't usually be achieved in a few one or two trials. Very high signal-to-noise ratio, and it is necessary to repeat the test several times, and then average processing can achieve a good control signal f. On the other hand, the complex information contained in the normal operation of the brain does not conform to such a simplified model, and such analysis will waste a lot of meaningful information, which is tantamount to buying and returning. Thus, the right month b separates the complex brainwave signals captured, and the non-meaningful noise is eliminated. All the research and analysis of the brain's physiological operations will be smoother. Furthermore, if the separated components can be compared with the previously obtained event data to confirm the occurrence of certain known events, the control interface between the human and the control device will make the manipulated mechanical device more operable. The operation of the mind, such as the brainwave control aircraft imagined in the movie "Firefox" will gradually be realized, and the 20 people who are limited by nerve transmission disorders (such as gradual rolling, etc.) will also be able to use brain waves. Control surrounding items and improve your self-care ability. SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a method for automatically analyzing and classifying a brain wave signal component. Another object of the present invention is to provide a brain wave signal 10 1274269 which can be said to be deducted. [A brief description of the main components of the drawings] 1 ••... The signal measuring device 22...the decomposition unit 2... ····Signal processing device 23...Selection unit 3 ... Storage device 24 ... Recombination unit 4 ••... Control device 25 ... ... classification unit 5 ·······User 251... Computer 11 ••...·measuring electrode 252...classifier 12 ...····sensor 41... drive controller 21 ••... seven processing unit 42 ... drive motor twenty one