algorithm - Comparing 2 one dimensional signals -


i have following problem: have 2 signals on time. same source should same. want check if are.

complications:

  • they may measured different sample rates
  • the start / end time not correlate. measurement not start @ same time , end @ same time.
  • there may time offset between 2 signals.

my thoughts go along fourier transformation, convolution , statistical methods comparison. can post me links can find more information on how handle this?

you can correct phase shifting them centers of mass line up. (or alternatively, in fourier domain multiplying inverse of phase of first coefficient.)

similarly, if want line images given partial data, can cross correlate , take maximal value (which again easy in fourier domain).

that leaves tricky part of process dealing sampling rates. if know a-priori sample rates are, (and if related rational number), can use sinc interpolation/downsampling rescale them common sampling rate:

https://ccrma.stanford.edu/~jos/st/bandlimited_interpolation_time_limited_signals.html

if don't know sampling rate, may bit screwed. technically, can try brute forcing on different rescalings of signal, doing tends either slow or else give mediocre results.

as last suggestion, if want match sounds can try using cepstrum , verifying peaks of signal close enough within tolerance. type of analysis used lot in sound , speech recognition, refinements make operate bit more locally. tends work best frequency modulated data speech , music:

http://en.wikipedia.org/wiki/cepstrum


Comments

Popular posts from this blog

c++ - Is it possible to compile a VST on linux? -

java - Output of Eclipse is rubbish -

jquery - Confused with JSON data and normal data in Django ajax request -