-----------------------------
Version 3.9:
* support full covariance for 'gmmp' command.
* add -B, -c1 and -c2 option to 'gmmp' command to specify block size and structure of covariance matrix.
* change temporary file path for csh.
* bug fixes.
Version 3.8:
* add 'c2ndps' command for transform from cepstrum to negative derivative of phrase spectrum.
* add 'mgclsp2sp' command for transform from LSP to spectrum.
* add 'ndps2c' command for transform from negative derivative of phrase spectrum to cepstrum.
* add 'wavjoin' command to save stereo wav file.
* add 'wavsplit' command to load stereo wav file.
* add -V option to 'dtw' command for loading DTW results.
* add -e option to calculate inverse matrix stably for 'delta' and 'vc' commands.
* support MAP estimation in 'gmm' command.
* replace -i option to -q option for 'lsp2lpc,' 'lspcheck,' and 'mgclsp2mgc' commands.
* change -c option of 'mlsacheck' command.
* update swipe and hts_engine API.
* bug fixes.
Version 3.7:
* add 'vc' command to perform GMM-based voice conversion.
* add 'lsp2sp' command to transform LSP coefficients to spectrum.
* add -R option to 'mlsacheck' command to specify threshold value.
* add -f option to 'gmmp' command to specify full covariance matrix.
* add -M, -c1 and -c2 option to 'gmm' command to specify block size and structure of covariance matrix.
* add -n option to 'delta' command to specify order of regression polynomial.
* add -magic and -MAGIC option to 'rmse' command to delete and substitute magic number.
* add -n and -t option to 'rmse' command to specify order and number of vector, respectively.
* add -POWX, -LOGX option, and new constant (semitone) to 'sopr' command.
* add -l option to 'decimate' and 'interpolate' command to specify length of vector.
* add -g and -G option to 'lspcheck' command to check value of gain.
* add -L option to 'lspcheck' command to specify log gain.
* add -n option to 'mgc2mgclsp' command to specify split number of unit circle.
* add -p and -d to 'mgc2mgclsp' command to specify maximum number and end condition of interpolation.
* add -x 16, 22, 24 and 48 option of 'glogsp' and 'grlogsp' command to specify more x-axis scales.
* add filter coefficients for 7:4 down-sampling, 3:5 and 7:8 up-sampling.
* unify option to specify log gain into -L option.
* fix -c option of 'mlsacheck' command to specify the type of modification of filter coefficients.
* bug fixes.
e8b440d001