Re: Slow Speech Playback


Chris P. [MVP] wrote:
<snip>

It can crudely work for some compressed formats, but you obviously can't
use the waveform similarity method so you'll get nasty clicking at each
block boundary. Some compressed formats will lose several milliseconds at
the beginning of each block if the previous block is missing, usually
reconvergent algorithms like ADPCM suffer this.

I see some test code development coming my way <g>

Isn't it always the way? :) I don't think you're going to have much other
option than to decode to alter speed smoothly. But if it's for playback
why do you have to recompress? Are you attempting server side manipulation
streaming to a client?


Actually 11025KHz GSM 610 works quite well, only quite minor artifacts.
I take the GSM data as a series of 65 byte blocks - ABCDEFGHIJK etc and
string them together as AABBCCDDEEFFGGHHIIJJKK etc, change the "data"
length to twice what it was - and bingo.

I might try assembling the blocks as AABCCDEEFGGHIIJKK etc for a 1.5
times slower speech.

Each block is of course 29mS of speech

Thanks for your input though, its always nice to get input from expert
others..

I'm writing for a "dictation to typist" system and this is for a
typist's utility to help decipher "strangled" speech.

Alan Lloyd

.