1

I'm working on a program that shall record audio recognizable, so it can later be easily compared with other audio files. The audio files will contain something like speech, so I was wondering what would be easier to do:

  • Implementing an algorithm for speech recognition and saving/comparing that outputs,
  • or implementing / creating an algorithm that creates something like a audio-fingerprint with e.g. Fast Fourier Transform and compares those?

Has anyone some experience on that area? I'm wondering if the second solution would be realizable within a relatively short period of time. Maybe there is a solution that is less hard to code and I'm just not finding it?

Andrew Thompson
  • 168,117
  • 40
  • 217
  • 433
Lejon Doering
  • 51
  • 3

1 Answers1

1

take a look at acoustic finger printing, most of the code you can find at github! it should help you solve your problem, more specifically the https://github.com/lalinsky/chromaprint/tree/master/tools

for more information take a look at this thread, this topic has already been extensively covered, Open source audio pattern recognition (finger printing) and Creating custom voice commands (GNU/Linux)

Community
  • 1
  • 1
bnm
  • 108
  • 6