Skip to main content

Table 6 RMSE of syllable logF0 contour (spn), syllable duration (sdn), syllable energy level (sen), and syllable-juncture pause duration (pdn) for the proposed approach and the baseline systems with various logF0 codebook

From: A parametric prosody coding approach for Mandarin speech using a hierarchical prosodic model

(a) SD case
  Inside (TrainTB) Outside (TestTB)
spn(logHz) sdn(ms) sen(dB) pdn(ms) spn(logHz) sdn(ms) sen(dB) pdn(ms)
 SD-HPM .070 4.8 .68 41.4 .064 4.7 .70 34.3
 SD-BSL-24 .069 5.0 .64 33.5 .066 4.7 .59 32.9
 SD-BSL-32 .065 5.0 .64 33.5 .061 4.7 .59 32.9
 SD-BSL-64 .053 5.0 .64 33.5 .050 4.7 .59 32.9
 SD-BSL-128 .044 5.0 .64 33.5 .042 4.7 .59 32.9
 SD-BSL-256 .037 5.0 .64 33.5 .042 4.7 .59 32.9
(b) SI case
  Inside (TrainTC2) Outside (TestTC)
spn(logHz) sdn(ms) sen(dB) pdn(ms) spn(logHz) sdn(ms) sen(dB) pdn(ms)
 SI-HPM .065 9.3 .80 44.8 .056 7.5 .66 44.9
 SI-BSL-10 .063 9.1 .78 42.0 .060 10.9 .88 39.4
 SI-BSL-16 .056 9.1 .78 42.0 .054 10.9 .88 39.4
 SI-BSL-32 .047 9.1 .78 42.0 .046 10.9 .88 39.4
 SI-BSL-64 .040 9.1 .78 42.0 .039 10.9 .88 39.4
 SI-BSL-128 .034 9.1 .78 42.0 .033 10.9 .88 39.4
 SI-BSL-256 .029 9.1 .78 42.0 .029 10.9 .88 39.4