I want to take other acoustic measurements (voicing, F0, RMS, tilt) at the same time as the frames of my LPC coefficients. Is there a way to extract the exact times that Praat is using to find the LPC coefficients? The time step is set at 10ms frames with 5ms overlap, so I guess I could just write a for loop as well, but I'm not sure how that would work the leftover time (if the sound isn't divisible by 5). I've also looked at the long and short text files, and I can't find anything that would indicate the times of these frames.
1 Answer
I believe that the answer is, "it's about in the middle". The LPC listing gives you a value for x1 which might be 0.02548848046722024, given a selected window length of .025. xmax might be .800962012144737 and should be the total length of the selection. With 151 reported frames and a time step of .005, that .755 sec. is 0.04596201214 less than the total duration. Given an offset of x1 by 0.02548848046722024, that implies that there is an un-analyzed tail of about 0.02047353167. So frame n starts at (n*step)+x[1]. I do not know why x1 is not offset a little less, so that the bits at the beginning and end are not more equal. Perhaps the best way to look at this is that there is a limit to the precision of your time computations: you could ask the authors.
