有別於穩態視覺誘發電位以定頻弦波為大腦人機介面的誘發訊號，Chirp視覺誘發電位以chirp訊號作選項閃光依據，本論文利用chirp訊號的頻率會隨時間線性變化的特性以及各選項閃爍時間與休息時間的交錯，設計變頻相位編碼的大腦人機介面。 Chirp視覺誘發電位使用的分析方法為分數傅立葉轉換，為了讓轉換公式在實現上更加有效率，本論文使用快速傅立葉轉換來實現分數傅立葉轉換以簡化實現的複雜度。 為了使視覺誘發電位的數位濾波器不需因為chirp訊號的變頻特性而使用過寬的通帶，本論文提出一方法來實現可線性調變通帶的帶通濾波器，其核心是使用分數卷積，達成在不同時間有不同通帶的數位濾波器並且能與chirp訊號的瞬時頻率匹配，此方法能提高系統的訊號雜訊比，進而提高大腦人機介面的辨識正確率。;This study implements a brain computer interface which is based on Chirp visual evoked potential. Because Chirp-VEP is not a sinusoid but a chirp signal, the phase-shift coding approach in this study would let each command has various instantaneous frequency at any time. Fractional Fouier Transform can analyze the characteristic of linear frequency modulation, but its computation takes a lot of time. A mothed in this study implements FrFT via FFT so that the complexity can be reduced. Unlike the unsuitable passband on a normal band-pass filter, the frequency-modulated passband can match the frequency of Chirp-VEP with time. It means that the filter completed by Fractional Convolution in this study has higher Signal-to-Noise Ratio than traditional filter. The experiment results demonstrate that the Chirp-VEP induced by the proposed method do have great accuracy and can be used in the BCI system.
