寬頻音訊編碼之聽覺遮蔽效應探討

Abstract

一種使用人耳遮蔽效應，來降低位元傳輸率的新型音訊編碼器，已在最近 幾年付諸實行。 編碼雜音，會被隱藏在臨界遮蔽值之下而無法察覺。這 是藉著運用一個心理聲響模型來做最佳的位元分配而達成的。然而，大部 分此型的音訊編碼器，如次頻帶編碼或轉換編碼，都是在頻域上做處理， 而顯示了非常複雜運算量的缺點。為了避免此缺點，在本論文中提出一個 『臨界遮蔽值適應性知覺濾波器』來搜尋『分析-利用-合成型編碼器』的 最佳激勵。此外，我們利用人耳主觀的聽覺特性，發展出一個一般化巴 克(Bark)頻譜失真測量值，來客觀地評估寬頻音訊編碼信號的品質。實驗 結果顯示，的確比其他傳統的客觀量測方法更為可靠。 種典型估計臨界 遮蔽值的方法，並用來計算音訊信號的論邊際位元傳輸率。 A new class of audio coding which exploits humaneffect to reduce transmission bit rate iscent years. The coding noise should be embedded under the just noticeable masking threshold to result in an spectrum. This can be accomplished by taking a model into account to optimize the bit-allocation However, most of audio coders focus on frequency domain approaches such as subband coding and transform coding exhibit the disadvantage of excessive computational load. In this thesis, we explore the benefits of ald adapted perceptual filter for excitation search in synthesis coders. We also develop an objective measure of quality, called generalized Bark spectral distortion to explore subjective perception of human ear. It has ated more reliable evaluation results than conventional ive measures do. Furthermore, some typical approaches for mating the masking threshold are studied to calculate cal bit-rate bounds for audio signals.