類神經模糊系統是由類神經網路與模糊理論所結合而成。類神經網路也稱為人工神經網路，主要模仿生物的神經系統，是一種平行計算系統，使用大量的人工神經元，來模仿生物神經網路的能力。目前視覺與聽覺方面表現最為出色的為生物機器，不管在圖像判別，或是語音辨識的能力，遠遠優秀電腦許多。而類神經網路主要就是利用生物神經網路的計算原理，來設計出一個功能強大的計算系統。模糊理論主要是描述的模糊性乃是隸屬程度上的不確定性，藉由隸屬函數 (Membership Function) 來表達。現今模糊理論被廣泛的利用在各領域中，其主要原因是透過模糊理論所設計的模糊邏輯控制器，設計理念貼近人類的思維模式，它是利用語言變數（Linguistic Variable）來代表以往的經驗與專家的建議，模擬人類對控制機械的經驗或操作行為，然後經由模糊推論工場（Fuzzy Inference Machine）模仿人類下決策的方法，將這些條件式控制規則轉化成自動控制策略，以達到控制目標的一種控制器設計方法。由於類神經網路優越的學習能力與判斷能力上，近年來被廣泛的應用在數位訊號處理上，在加上模糊理論對於未確定、部分未知的變化等，具有其推論效果。在本論文中，利用類神經網路針對多訊號源作數位訊號處理，再利用模糊理論，把其眾多結果利用模糊推論方式，推論出較合適的訊號可能性。本論文所判斷資料為自行產生訊號組合，利用常見的函數種類合成，例如：正弦波、餘弦波、方波、波物線等。本論文在最後呈現實驗結果，本機制在第一順位之預測結果，分別為 73.33%、83.33% 以及 83.33%。在前三順位之預測結果，分別高達90.00%、100.00% 以及 96.67%。證明在分類預測上，本論文之機制能夠提供更準確之預測。The Neural Fuzzy System is a combination of Neural Network and Fuzzy Theory. Neural network is also known as artificial neural network. It mimics the biological nervous system. It is a parallel computing system. It uses a lot of artificial neurons to mimic the biological neural network. In the artificial intelligence field, Neural Networks have been applied successfully to speech recognition, image analysis and adaptive control. The Neural Network is using the calculation principles of biological Neural Networks to design a powerful computing system.Fuzzy Theory is a form of many-valued logic or probabilistic logic. It deals with reasoning that is approximate rather than fixed and exact. The Fuzzy Theory devotes the main of the fuzzy membership degree of uncertainty. Fuzzy Theory is widely utilized in various fields. The main of reason is the fuzzy logic controller designed by Fuzzy Theory. The concept closes to the human mode of thinking. It is using linguistic variables to represent the past experience and the advice of experts. By Fuzzy Inference Machine, it mimics the human’s decision. In this paper, it is using Neural Network to deal with multi-source. By Fuzzy Theory, the numbers of Neural Network’s results are to combine to form a new result. It judges the self-generated signal combination. It uses the common of types of functions, such as sine wave, cosine wave, and square wave. Finally, the simulation results show that the proposed can accurately predict on top 1 by 73.33%, 83.33% and 83.33%. It can accurately predict on top 3 by 90.00%, 100.00% and 96.67%. This proves the proposed mechanism can provide more accurate predictions.
