對MPEG環場音訊壓縮子頻-時間處理之改善

Abstract

MPEG Surround Audio Coding是一個低位元率多聲道音訊壓縮的標準。其壓縮的概念是利用降混音(down‐mix)的處理和計算空間參數(spatial parameter)的方式來減少聲道數。而解碼端則是將耦合訊號經過去相關器(decorrelator)的處理，所產生出的去相關訊號再根據空間參數進行升混音(up‐mix)處理來重建聲源定位與空間寬廣度的環繞效果。但訊號在經過去相關器的處理後，在暫態(transient)處會被破壞造成感官聽覺上的影響。MPEG Surround為此提供了Subband Domain Temporal Processing (STP) 工具來解決這個問題，但就客觀測試和主觀測試後的結果發現，經STP處理後反而都是比較差的情況。 此篇論文會針對STP在做法上進行一些修改，藉此來改善在主觀測試和客觀測試上的結果。

MPEG Surround Audio Coding is an efficient compression standard of high-quality multichannel sound. The compression concept uses down-mix and spatial parameters calculation to reduce the channel number and achieve high compression rate. In decoder, an upmixing module can reconstruct the multichannel signals by using the output signal of decorrelator and the spatial parameters. After the signal is processed by decorrelator, the transient property has been destroyed, affecting the perceptual listening quality. In order to overcome the defect, MPS proposes Subband Domain Temporal Processing (STP) shaping tool. However, the objective and subjective quality assessments still show negative results on the STP-processed signals. This thesis modifies the current STP structure for improving the results of the quality assessments