近年來類比式ANC耳罩已經成為蠻普遍的產品，但是不同的耳罩需依其規格來設計其控制器，而且此類比控制器的階數通常非常的高而導致難以實現。因此本論文採用適應性控制器來發展ANC耳罩，此種方法可以適用於市面上的耳罩來到消音的目的。 本研究基於DSP卡，在前饋、回饋及複合三種架構上實現FXLMS演算法，並且使用此三種架構比較消音的效果以及了解其特性。相較於其它架構，複合架構在有干擾噪音的情況下的確有很明顯的優點。接著，製作耳道模型模擬人耳聽覺感受到消音的情形，並且考慮誤差麥克風至鼓膜的距離，希望能應用虛擬麥克風的技術讓人耳能感受到最大的消音。除此之外，本論文中也探討估測路徑階數以及採樣頻率對ANC系統的影響。 Recent years, ANC headsets with analog controllers have become popular products. But the analog controller must be designed according to different headsets for specific tasks, and the order of the analog controller is usually too high to realize. In the thesis, digital adaptive controllers are realized and adopted for the developed ANC headset. The design architecture can be applied in commercialized headsets for noise-cancellation purpose. In the study, three control structures, i.e., the feedforward, feedback, and hybrid ones with the FXLMS algorithm, are investigated to design the controller based on a DSP broad. The performance and characteristics of noise-cancellation using different ANC structures are compared. The hybrid structure superior to the others is highlighted. Then, an auditory canal model is built up to evaluate the result of noise reduction in human hearing. Considering the actual difference between the error microphone of the headset and human eardrum, the virtual microphone technique is applied. Besides, the influences of the order of the estimated paths as well as the sampling rate on the ANC system are investigated.
