- 學位論文
結合P300與SSVEP之即時BCI居家照顧系統
A Real-Time Brain-Computer Interface home care system combining P300 and SSVEP
摘要
摘要 近年來，全球邁入高齡化的社會型態，而長期臥床的病患或有運動障礙的病人，諸如：中風、行動不便的病患等等逐年增加，利用腦機介面(Brain-Computer Interface, BCI)系統提供了另一種與外界溝通的管道，這種溝通管道對於此類病患有著相當大的貢獻及幫助。 本論文設計一套結合P300與SSVEP的居家看護系統。此系統不需要看護人員在一旁協助操作，病患就可自行完成，病患可透過注視視覺刺激面板刺激誘發SSVEP電位，藉由此訊號判斷病患目前是否想起動面板，在啟動面板後所注視的25個看護功能可達到與外界溝通的一種方法，這樣可以比拼字來得有效率，病患有問題可以馬上通知家人、醫生或是護士目前的生理狀況，並且將需要求助的資訊傳送到看護人員的通訊系統上。 本論文設計一套居家照顧面板，能讓患者透過視覺刺激傳達到控制面板上所顯示的居家照顧需求，此面板是由Visual C#所開發。系統啟動後，會出現以10Hz頻率閃爍的start字樣，並利用了穩態視覺誘發電位(Steady State Visually evoked Potential, SSVEP)方法，經由快速傅立葉(Fast Fourier Transform, FFT)將腦波轉換至頻率域，接著計算頻帶功率(Band Power, BP)，最後判斷Band Power是否達到啟動下階段條件(Threshold)，如此一來在面板開啟後，利用SSVEP特性使面板不會隨意進入25個照顧功能閃爍階段，好處在於面板控制是由患者自行決定，並不需照護人員在一旁操作。 當患者注視start字樣並且成功觸發啟動後，系統將進入25個居家照顧面板功能階段，這時25個照顧功能將以行列隨機不重複閃爍，系統會利用視覺刺激所誘發出的腦波事件相關電位，經由資料的處理後，將腦波訊號於閃爍刺激後500豪秒做切割，並將多次刺激的腦波訊號經過疊加平均的處理，此方法能突顯P300事件相關電位以及增加訊號雜訊比，再經由支持向量機(Support Vector Machine, SVM)演算法辨識，判斷其行與列是否有出現P300事件相關電位。系統將辨識後的結果呈現在面板上方空格處，接著會出現以10Hz頻率閃爍的send字樣，再利用SSVEP的特性，判斷Band Power是否達到啟動下階段條件(Threshold)，當患者注視send字樣並且成功觸發啟動後，最後將結果以e-mail方式發送到看護人員的通訊系統或是手機。
並列摘要
In recent years, due to aging problems emerging in social societies worldwide, the number of long-term bedridden patients or patients suffering for severe mobility impairments caused by reasons such as stroke and motor disabilities are expected to rise. BCI (Brain-Computer Interface, BCI) systems can have great contribution to such patients, since they can provide a channel to communicate with the outside world. In this thesis, we designed a resident care system based on a combination of P300 potential and SSVEP. The system can be operated by the patients alone, requires no assistants from others. SSVEP signal is induced when the user is looking at the resident care panel, which is then used to determine whether the user wants to activate the panel. Communication with the outside world can be achieved by looking at 25 resident care functions, which is more efficient compared to spelling, the patient can instantly notify his family、medics or nurses of his physiological state if there is anything wrong, and send help messages to the care taker’s communication system. The resident care panel was developed with Visual C#, and the user can choose which resident care function presented on the panel to be sent out by looking at the panel. After the system is launched, SSVEP (Steady State Visually Evoked Potential, SSVEP) signal is induced when the user is looking at a “start”, flashing at a frequency of 10 Hz, later the signal is transformed into frequency domain by performing Fast Fourier Transform (FFT) and its Band Power (BP) is calculated. Given a threshold, the panel activates only if the Band Power exceeds it, preventing the panel from entering the next stage, the flashing stage, after the system is launched, so that the user can control the panel without others’ assistance. When the user successfully activates the panel, the system will enter the resident care function stage, these 25 functions will flash in a random, non-repeatedly, column and row manner. The system will further process the event-related potential induced by the visual stimulus, then cut the event-related potential into a 500 ms segment after the visual stimulus appears, and signal-to-noise ratio is lowered and P300 event-related potential is highlighted after segments undergo linear superposition. Classification is done by Support Vector Machine, detecting which row and column induce P300 event-related potential, and the system will present the result on the top of the panel, and a “send” will flash at a frequency of 10 Hz, using the characteristic of SSVEP, if the Band Power exceeds a given threshold, the result will be delivered by e-mail to the communication system or mobile phones of nursing workers.