[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:03969] problems with sub-band aperiodicity


Dear all,
I extracted aperiodicity (ap) with Straight, then I used the script to convert them to sub band ap:

echo "Converting aperiodicity file cmu_a0003.ap to band aperiodicity file cmu_a0003.bap"; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L  64 -s   0 -e  63 -S 0 | ./average -l  64 > bap1; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L  64 -s  64 -e 127 -S 0 | ./average -l  64 > bap2; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L 64 -s 128 -e 191 -S 0 | ./average -l 64 > bap3; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L 64 -s 192 -e 255 -S 0 | ./average -l 64 > bap4; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L 64 -s 256 -e 319 -S 0 | ./average -l 64 > bap5; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L 64 -s 320 -e 383 -S 0 | ./average -l 64 > bap6; \
   ./x2x +af cmu_a0003.ap | ./bcp +f -n 512 -L 129 -s 384 -e 512 -S 0 | ./average -l 129 > bap7; \
   ./merge -s 0 -l 1 -L 1 bap1 bap2 | \
   ./merge -s 2 -l 2 -L 1 bap3 | \
   ./merge -s 3 -l 3 -L 1 bap4 | \
   ./merge -s 4 -l 4 -L 1 bap5 | \
   ./merge -s 5 -l 5 -L 1 bap6 | \
   ./merge -s 6 -l 6 -L 1 bap7 > cmu_a0003.bap;

Then I used the script to regenerate the ap from bap file before synthesizing speech with new ap file.

./bcp +f -l 7 -L 1 -s 0 -e 0 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 64 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap1; \
   ./bcp +f -l 7 -L 1 -s 1 -e 1 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 64 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap2; \
   ./bcp +f -l 7 -L 1 -s 2 -e 2 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 64 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap3; \
   ./bcp +f -l 7 -L 1 -s 3 -e 3 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 64 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap4; \
   ./bcp +f -l 7 -L 1 -s 4 -e 4 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 64 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap5; \
   ./bcp +f -l 7 -L 1 -s 5 -e 5 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 64 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap6; \
   ./bcp +f -l 7 -L 1 -s 6 -e 6 -S 0 cmu_a0003.bap | ./dfs -b 1 -1 | ./interpolate -p 129 | ./dfs -a 1 -1 > ./ap/cmu_a0003.ap7; \
   ./merge -s 0 -l  64 -L 64 ./ap/cmu_a0003.ap1 ./ap/cmu_a0003.ap2 | \
   ./merge -s 128 -l 128 -L 64 ./ap/cmu_a0003.ap3 | \
   ./merge -s 192 -l 192 -L 64 ./ap/cmu_a0003.ap4 | \
   ./merge -s 256 -l 256 -L 64 ./ap/cmu_a0003.ap5 | \
   ./merge -s 320 -l 320 -L 64 ./ap/cmu_a0003.ap6 | \
   ./merge -s 384 -l 384 -L 129 ./ap/cmu_a0003.ap7 > ./ap/cmu_a0003.ap; 

The problem is that: the synthetic speech is very different from the original speech. I don't know where I am wrong. Can you give me some advices. Many thanks. I attached the original and the synthetic speeches.

Best regards,
Dinh Anh Tuan

Attachment: cmu_us_arctic_slt_a0003.wav
Description: Wave audio

Attachment: cmu_a0003-2a.wav
Description: Wave audio