This paper proposes a tonal component coding algorithm for a codec that emp
loys a transform followed by Huffman coding, such as MPEG-2 Audio AAC (adva
nced audio coding) [1].
After the input audio signal is mapped onto the frequency domain, the propo
sed algorithm removes local maxima components that degrade the coding effic
iency. As a result of this withdrawal, the flatness of the spectrum increas
es and the efficiency of Huffman coding is improved. The removed components
are encoded separately as side information.
When the frequency resolution of the time/frequency mapping is high, this a
lgorithm works more effectively since local maximum samples appear more fre
quently with such a mapping. Simulation results show that this algorithm ac
hieves as much as an 11% bit reduction per frame and improves the coding ef
ficiency in 41% of all audio frames. (C) 1999 Scripta Technica, Electron Co
mm Jpn Pt 3, 82(4): 71-78, 1999.