This paper provides a subjective quality analysis of transforms used in aud
io compression algorithms for a class of music signals. A 34-subject listen
er test compares three transforms in conjunction with an MPEG I layer 1 com
pression scheme. One test compares the performances of the discrete wavelet
packet transform (DWPT) and the modified discrete cosine transform (MDCT)
used in MPEG. Another test compares the performances of a DWPT eight-level
nonuniform critical-band split and a DWPT five-level uniform subband split.
Results indicate that the critical-band split provides significantly bette
r quality than the uniform subband split for sounds with tonal and strong l
ow-frequency content, while the DWPT outperforms the MDCT with significant
improvement for nontonal sounds. (C) 2000 Acoustical Society of America. [S
0001-4966(00)03907-2].