本論文裡我們使用獨立成份分析法(ICA)和減法聚類分類能夠有效率的分析老鼠的腦皮質層電生理訊號(ENG,腦神經電圖)。系統中我們發展一套完整的系統能夠紀錄八通道的腦神經電圖,此八通道微電極陣列被植入實驗中老鼠大腦皮質層內的主運動區,訊號經由前級放大電路與後級帶通濾波放大電路來放大,並透過A/D 卡把腦電生理訊號存取到電腦。當紀錄的同時,系統同步攝影監控紀錄老鼠的動作。紀錄腦神經訊號之後,我們使用ICA方法分析最可能獨立的原始腦電生理訊號成份。但是所得到的獨立原始訊號成份會有幾個問題存在,其獨立原訊號成份的排序、比重和大小都不能被確定,因此我們發展了源訊號關聯匹配法來解決此問題,並且使用減法聚類分類法來分類老鼠的腦神經電訊號。從實驗結果中已我們能夠有意義的分辯出老鼠做動作時與腦波訊號之間的關係。
In the study, we propose an effective method using ICA and subtractive clustering to analyze (Electroneurography) ENG signal of the rat. We have completed the system which can record 8-channel ENG signal. Using the 8-channel multi-electrode array and then implant it into the primary motor cortex of rats. The signal is amplified through front-stage amplitude circuit and after-stage amplitude band-pass filter circuit. And, we record the ENG signal into PC through A/D card. When recording the ENG signal, we also synchronously monitor the activity images of rats at the same time. After recording we employ to ICA method, we get independent sources as possible as we can. But the independent sources still have some problems. It is difficult to determine the order, weight, and the scaling of independent sources. We develop a matching source method to solve this problem, and we used the subtractive clustering which can classify the ENG signal of rats. The result we can meaningfully distinguish relation between the ENG signal and the activity of rats.