智慧型揚聲器非線性診斷系統

Abstract

本專利目的在建立一個智慧型揚聲器失真診斷專家系統。揚聲器工作在大訊號的情形下，許多揚聲器部位的非線性效應便會顯現，非線性會造成聲音的失真(Distortion) 破壞音質。過往針對揚聲器缺陷的診斷仰賴專門人士進行人工判讀，在效率、方便性、一致性上均不及電腦。本系統使用模糊類神經網路架構 (Neural Fuzzy System)製作專門用於揚聲器缺陷診斷的專家系統，針對不同的失真型態，可以診斷出造成失真的來源並提出改善建議。文末，兩個真實的揚聲器將作為實驗品，比較此專家系統與傳統方法的診斷結果。結果可顯示，兩者診斷結果吻合，且專家系統較傳統診斷方法更為簡單、更有效率。

An intelligent diagnostic system for the voice-coil loudspeakers is proposed. The defects discussed in this paper are caused by loudspeaker nonlinearities. When a loudspeaker works in the large signal domain, many nonlinearities of the loudspeaker may appear, and nonlinearities cause distortions destroying the sound quality. This intelligent diagnostic system consists of a defect database and a fault inference module. There are six kinds of defects chosen in this paper, and a large signal modeling technique using the electromechanical analogous circuit is employed for generating the defect database. A neural fuzzy network is exploited for intelligent inference of defects base on the defect database. Experiments of two real loudspeakers indicate the reliability of this diagnostic system.