為重度肢體障礙患者設計之腦機介面

Design of Brain Computer Interface for the Person with Severe Physical Disability

林清助(Ching-Zhu Lin)

指導教授 ： 吳崇民

崑山科技大學/工程學院/電子工程研究所/碩士(2009年)

摘要

隨著科技的進步，人們的生活越來越多采多姿，在食衣住行育樂等方面都有科技來滿足人們的需求，但對於身心障礙團體來說，相對的輔助設備，卻明顯不足。尤其是重度肢體障礙患者，如脊髓損傷、運動神經元疾病患者等長期癱瘓在床的病人，對於溝通和行動障礙兩者並存的患者而言，其日常生活起居處處充滿著不便，一般與此類患者溝通最簡單的方式就是製作一些簡易讀字卡，依著基本的身體或心理層面去推敲他的需要，如天氣轉涼了，「冷」「熱」「穿衣服」等…，但這並不能完全表達患者的需求，因此常造成無法滿足患者需求的情況，造成生活上非常的不方便。
本研究透過腦波儀擷取大腦腦波，以建立實用性高的即時腦波摩斯碼文書輸入系統(EEG Morse code Text Input System , EEG McTin)為前提，作為重度肢體障礙患者及照護者之間的橋樑，透過特定動作(呼吸、下顎動作) 腦波的差異性，搭配g.tec所開發的腦波記錄儀擷取腦波訊號，經過腦波人機介面做訊號處理後，轉換為摩斯碼訊號，再透過摩斯碼文書輸入系統(Morse code Text Input system, McTin)辨識後，轉換為可讀字元、電腦操作指令或家電控制訊號。
在現今研究中所採用的均為多點量測，在裝置上不易且訊號分析困難，導致辨識速度慢，本研究主要使用單點量測，減少以往研究中裝置不易以及訊號分析困難的問題。而在腦波訊號轉換部分主要使用史密特觸發來進行訊號轉換，且透過模糊史密特觸發演算法來調整判斷準位，達到高適應性的目的，系統可依據使用者當天狀態可透過自動準位調整，達到良好的辨識結果。此外，傳統摩斯碼輸入方式在長短音必須以3:1的比例輸入，而間隔音也是3:1，這顯示了使用者必須保持穩定的輸入速度，增加了摩斯碼使用上的困難，然而McTin系統主要採用長短音分離追蹤模糊演算法(Long-Short Separation Fuzzy)處理傳統摩斯碼的比例問題，讓使用者可以輕鬆的輸出想要的字元或命令，並降低輸入摩斯碼的困難度及限制。
本研究可在每0.125秒可下達一個控制命令(如ON/OFF開關命令)，若是複雜指令(如打字)，約3~5秒鐘可以完成。第一次使用的受測者在使用McTin訓練系統進行30分鐘的訓練後，受測者平均可在55.33秒內完成摩斯碼A~J的練習，其平均辨識率可達到92.57%，在訓練二個禮拜內受測者平均可在37.59秒內完成摩斯碼A~J的練習，平均1個英文字母約為3.76秒左右，其平均辨識率可達到94.7%，使用者可在短時間內熟悉並操作EEG McTin系統。
以往研究中常發生腦波辨識速度慢或辨識率低導致難以實現腦波即時控制系統，而在訊號擷取時，又常因偶發之肌電或眼動訊號干擾，造成辨識錯誤影響系統的操作性及可信度，本研究改善這些問題提供一個腦波即時輸入溝通系統，讓患者不需要依靠周邊神經和肌肉等，透過功能健全的大腦就可以達到與外界溝通、傳達訊息、以及自主行動和自我照顧等功能。照護者可以藉由此系統了解重度肢體障礙患者的需求與想法，而患者亦可藉由本系統增進與周圍人事物之互動關係，提高生活品質。

關鍵字

EEG ； 腦機介面 ； 摩斯碼 ； 輔具

並列摘要

With the technologic advancement, people’s life become more and more colorful, there are many technology product to satisfy people’s requirement with their food, clothing, housing, traffic, education and amusement, but there and not enough supplementary instruments for disabled groups. Especially for severe extremities disabled such as Injury to spinal cord, Motor neuron disease and so on, who are paralyses in a bed for a long-term. The people who have communication and action disabled, they couldn’t do them well, so their lives are full of inconvenience. If they want to associate with the normal people, the most easy method is to make some word cards of reading. With these word cards to guess what he need from his health or psychology, such as the weathers turns cold, ”cold”, “hot”, “wearing clothes” and so on, But it couldn’t express all the thought from patients. Therefore, it often caused that we can’t know what the patients want or their condition. It’s too inconvenient for their lives.
In this research, through the instrument of EEG to establish high practicality of EEG Morse code Text Input System, EEG McTin, and it can become a bridge between severe disabled and caretaker. Through the differences in EEG of specific movement (breathing, mandible movement). With g.tec develops the instrument of EEG. After the signals process in process in Brain-Computer thansforms for the Morse code signals, and through Morse code Text Input System, McTin, it can transform the Morse code signals to character what we can read instruction code of computer and control signal of household electrical appliances.
In the current research and used in multi-point measurement, the device is not easy to analyze. The research measured main useful single point and reducing the problems of the device EEG signal in the main part of the conversion to the use of Schmitt trigger signals conversion, and through the fuzzy algorithm of the Schmitt trigger to adjust the position to achieve the purpose high adaptability, the system can adjust the automatic position to identify good situation on the same day. In addition, tradition Morse code input in the length of the long tones and short tones must be imported in the radio 3:1, and the tone spacing is 3:1, indication that the user must input speed to maintain stability, increasing the use of Morse code difficulties. However, the system of McTin is mainly used to Long-Short Separation Fuzzy algorithm to deal with the proportion of the traditional problem of Morse code so that users can easily output the characters or want to order, and reducing the importation of Morse degree of difficulty code and restrictions.
In this research, issue a control command in each of 0.125 seconds (such as ON/OFF switch command), if the complex instructions (Such as typing), completing about 3 to 5 seconds. People who were trained by first time in the use of McTin system after 30 minutes, they were completed to practice A to J of the Morse code about average 55.33 seconds, the average recognition can be achieved 92.57%. They were trained on two weeks in the use of McTin system, they were completed to practice A to J of the Morse code, about average 37.59 seconds, the average in an English word is about 3.76 seconds. It’s average on recognition may achieve 94.7%, the users may be familiar with EEG McTin system in short time.
In previous research often happened some problems, such as brain waves are too slow or low rate of recognition, it is difficult to control with control system. And when the signals are extraction, the signals are often disturbed by the accidental muscle electricity or the eye move. They can cause the wrong identification and influent the operation system and the confidence level. This research provides a brain wave real-time input communication system to improve these questions don’t need to depend upon the peripheral nerve and muscle so on, they may communicate with outside by the severe physical disabilities of the demand and idea by the system. The patients may also promote with the people who around them and improve the life quality by the system.

並列關鍵字

EEG ； BCI ； Morse Code ； Assistive

參考文獻

[4] Ramaswamy Palaniappan,“A New Brain–Computer Interface Design Using Fuzzy ARTMAP.” IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 10, NO. 3, SEPTEMBER 2002。

[5] 陳志瑋，“研究以小波神經網路作μ波即時鑑別”，碩士論文，國立成功大學，2002。

[9] 黃名斌，“USB介面之模組化腦波記錄儀”，碩士論文，中原大學，2003。

[11] Kjell Ohlsson，“A Neural Network Based Brain Computer Interface for Classification of Movement Related EEG.”，2003。

[12] Supratim Ray ，“High-frequency gamma activity (80–150 Hz) is increased in human cortex during selective attention.”，2007。