I encountered a problem when using 'lspdf' for synthesis:
When the lsp order is relatively large, e.g. m>20, some synthesized points' value turn out to be extremely large, which makes the synthesized speech sound annoying. And larger the lsp order, more the abnormal synthesized points.
Anticipate your answer about this question!
Have you ever tried to use 'lspcheck' command ?
The stability of LSP coef. can be checked.
Hi , I had used the lspcheck to check the stability of LSP coef. However, it seems not become good. I compared the file named $real_name.lsp and $real_name.rlsp, I found that they are the same.
My whole command is as following:
$tool_path = 'D:\SPTK\bin';
$frame_len = 512;#fram_length 25ms
$frame_shift = 80;#frame_shif 5ms
$sample_rate_k = 16;
$para_order = 40;
$warp_factor = 0.42;
system("$tool_path\\x2x +sf <$real_name.short | $tool_path\\pitch -a 1 -s $sample_rate_k -p 80 -L 110 -H 280 >$real_name.pitch");
system("$tool_path\\x2x +sf <$real_name.short| $tool_path\\frame -l $frame_len -p $frame_shift|$tool_path\\window -l $frame_len |$tool_path\\lpc -l $frame_len -m $para_order >$real_name.lpc");
system("$tool_path\\excite -p $frame_shift $real_name.pitch | $tool_path\\poledf -m $para_order -p $frame_shift $real_name.lpc >$real_name.syn");
system("$tool_path\\lpc2lsp -m $para_order -s $sample_rate_k -n 258 < $real_name.lpc > $real_name.lsp");
system("$tool_path\\lspcheck -m $para_order -s $sample_rate_k -r 0.01 < $real_name.lsp > $real_name.rlsp");
system("$tool_path\\excite -p $frame_shift $real_name.pitch | $tool_path\\lspdf -m $para_order -p $frame_shift $real_name.rlsp >$real_name.syn");
Using the same commands, I just change the $para_order from 40 to 20, the synthesized wave sounds good.But for 40-order, it become bad.
Sign up for the SourceForge newsletter:
You seem to have CSS turned off.
Please don't fill out this field.