本論文討論老鼠尾巴受到刺激時，其腦部感覺區域中多個神經元，所產生的動作電位樣式。分析的資料型態是將真實訊號的動作電位，轉譯為離散的時間序列，在統計上稱為點過程。對相同強度的外部刺激，尾巴產生不同程度的反應行為，透過單位時間動作電位發生次數的統計，進行訊號和反應行為間的關聯性分析。 針對常態下訊號的討論，發現感覺神經元本身並非穩定的點過程。於是進一步對不同的反應類型，分別探討刺激前和刺激後，單位時間動作電位發生次數的變化。觀察發生次數期望值的趨勢，分析不同反應行為的訊號特徵。另外，以刺激前後動作電位的數量差，觀察不利於分析的資料點及神經元，使得資料分析的特徵更顯著。 最後藉由討論得到的樣式特徵設計演算法，對訊號進行反應行為的分類測試。並由分類的結果，觀察刺激前後不同反應行為的神經訊號，找出特徵性最好的判斷範圍。以此驗證及說明，神經訊號與行為反應的關聯性。 This paper discusses the neuron in the rat brain which produces action potentials when the rate tail receives the stimulation. We translate the action potentials of real signal to the discrete sequence, which is called point process. The tail has different response to exterior stimulation of the same intensity. And we use histogram with di erent bin sizes to analyze the connection between the neural signals and the reactions. Considering the normal neural signals, we nd that the point process isn't stationary. So we further analyze the variation of the counts for action potentials before and after the stimulation for the di erent reactions. By observing the expected value of the counts, we characterize the pattern of different reactions. Moreover, the patterns can be made more notable if we observe the disadvantageous data and neurons by using the difference counts of spikes before and after the stimulation. Finally, we design an algorithm to classify the reactions with the neural signals. And by classified results, we observe the different response from stimulation and discover the better characteristic range. It can prove and explain the connections between the action potentials and the reactions.
