- 學位論文
基於獨立信號分析法利用時間延遲神經網路模擬老鼠大腦皮質層誘發實際動作的信號語言
Using Time delay Neural Networks to Simulate Signal Language of Action from Cerebral Cortex Evoked Potential of SD Rat Based on ICA
摘要
廣泛的電生理研究的單神經元 40 多年來積累了許多對皮質功能分區的知識，目前許多研究重點是了解神經元是否集體反應，不能直接推斷的反應單神經元的關係，目前尚未有完整的研究。本研究實驗，先取得老鼠皮質層之訊號並利用獨立成分分析法以獲得獨立來源訊號，應用倒傳遞學習時間延遲類神經網路，建立分類系統，假設大腦解讀神經元訊號的方式與人類解讀語言方式雷同，我們將訊號編碼成單字與字母並觀察編碼成果。藉由編碼各種動作的神經訊號，產生大量的編碼結果，利用編碼結果去分析皮質層訊號與大腦運作原理。本研究推斷神經元間反應關係，以了解腦皮質動作。
並列摘要
More than 40 years, widespread electrophysiological study of the single neuron has accumulated knowledge of cortical functions. Many studies focus whether it was collective response of neurons which couldn’t directly infer the relationship between single neuron that it’s not complete research. In the experiment, getting the signal of rat cortical layer and using ICA obtains an independent source signals and then apply the time-delay back-propagation neural network to establish a classification system. Assume that neurons in the brain interpreting signals is similar to human interpreting the language and then we code words into letters and observe the coding results. By coding nerve signals from various kinds of motion, it produces a large number of coding results and then we analyze the results to understand the operational mechanism between cortex signal and brain cortex. This study inferred response relationship between neurons in order to understand mechanism of cerebral cortex.