| Summary: | 碩士 === 國立臺灣科技大學 === 工程技術研究所 === 83 === This paper explores the application of subband coding techn
ique to high quality audio compression. When compressing audio
s ignals, coding error inevitably exist in the reconstructed
signa ls. To maintain the quality of the original audio
signals, it is necessary to make the coding error inaudible.
Thus, the masking e ffect of the human hearing characteristics
can be fully utilized to minimize the effect of coding errors.
In this paper, we close ly examine the masking effect and
judiciously apply it to the au dio signal compression. In the
audio signal coding technique, we will record the ma ximum
amplitude of each subband signal as side information. To r
educe the bit rate further, a common maximum value will be
chose n for a adjacent and closely correlated frames. Only a
single s et of maximum values are transmitted for the save
collection of subband signals, which significantly reduces the
amount of side information needed to be transmitted. To
minimize the quantizati on error, we adopt the u-law
quantization scheme in signals, qua ntization and the optimal
value of u is determined for a set of training audio signals.
|
|---|
