穩態視覺腦機介面於遊戲設計應用

Abstract

近年來隨著科技發展，腦機介面(Brain computer interface)技術日益精進，許多不同面向研究如雨後春筍般出現，所謂的腦機介面一詞，泛指使用頭部與事物控制之間的橋接，其技術包含移動想像(Motor Imagery)、視覺誘發電位（Visually Evoked Potential）、穩態視覺誘發電位（Steady-State Visually Evoked Potential）、P300波等，相較於其它系統，穩態視覺誘發電位具有較高的訊雜比、較高的處理速度、以及同時選擇大量的目標等優點，所以本研究使用了穩態視覺誘發電位作為開發。 以往大部分研究使用α(8~13Hz)及β(14~30Hz)頻段，並選擇黑及白交替做為刺激源，但在實際的真實世界中，事物往往是多彩的，因此本研究使用了近年來非常流行的三消遊戲(例:Candy Crush)做為範例，針對各閃爍頻段及時間長度，探尋穩態視覺誘發電位於實際使用時的最適合設計。實驗共收錄十位視力正常且無疾病的受測者，16個腦電位置(無前腦區)，其中O1、Oz、O2用於實現穩態視覺誘發電位，並以典型相關分析(Canonical correlation analysis)做為分析方式，其餘位置用於探討人們執行時的認知狀況。 根據分析結果，於θ(4~7Hz)頻段、閃爍時間5秒時，得到本研究中的最佳分類率(98.3±1.0%)，而進一步使用訊息傳輸速率(information transfer rate)評估效能，得到於θ頻段、閃爍時間2秒時，最高訊息傳輸速率為45.29±6.4(bits/min)。最後根據實驗結論，本研究結合了無線可攜式腦波偵測儀，設計了一款前置簡易且即時互動的三消遊戲，讓腦機介面實際應用於遊戲層面中，並可以使普羅大眾能更直觀的體驗成果。

Brain computer interface (BCI) is a medium of communication through which users can control a device or application. BCI has been gaining a lot of attention in recent years for wide range of application like robotics, rehabilitation, neuro-marketing, gaming etc. Motor Imagery (MI), Steady-state visually evoked potential (SSVEP) and P-300 are the frequently EEG signal components for application purpose. SSVEP based BCI systems have some advantages over other like high information transfer rate (ITR), high signal to noise ratio (SNR) and require little user training. Most of the previous research studies implemented a black/white stimulus and the frequency of the stimulus was in alpha/beta band. However, in real-world application the objects are colorful. Therefore, this work focused on exploring the performance different frequencies of colored stimulus. In total, 16 frequency stimulus (4 from each EEG band i.e., theta, alpha, beta and gamma band) were tested for developing a practical SSVEP based neuro-gaming system. For this, we used 16 channels (placed non-invasively on posterior region of the brain) to record offline data EEG; but only channels O1, Oz, O2 were used for implementing an SSVEP based BCI application and rest of the channels were included in exploring the human cognition. The stimulus was presented as match-three game with flickering gems similar to the popular Candy Crush game. Using Canonical correlation analysis (CCA) as classification algorithm, highest accuracy was obtained with theta band with a stimulus duration of 5 seconds. To further evaluate the performance, we estimated information transfer rate (ITR) and accuracy for different stimulus durations ranging from 2~5 sec. Highest ITR of 45.3 bit per min is achieved with theta band at 2 sec stimulus duration with an accuracy of 92.4%. Therefore, with this work, we found the best SSVEP based BCI system parameters that can be used for developing a practical neuro-gaming system.