-1
$\begingroup$

I have two waveforms, and I'm not sure which of them is for the word "cease" and which one for "shoot". Now I guess based on voiced and unvoiced waveforms(voiced should be periodic), the first one is "cease" and the other is "shoot". My question is, how to determine which waveform below refers to the word "cease"/"shoot"?

(1) enter image description here

(2)

enter image description here

$\endgroup$
  • $\begingroup$ And your question is ..... what? $\endgroup$ – JRE May 22 '15 at 15:02
0
$\begingroup$

it is almost impossible to see the different by 1000-5000 samples data point. If you have 5 seconds of signal with 8000 Hz sampling freq, it will give 40000 sample points and it will clear enough to differ cease/shoot. However, the acoustic signal processing community prefer to use spectrogram to check the spell.

This is the waveform of cease (left) and shoot (right), it is similar but easy to differ. Waveform of cease and shoot

And here is the spectrogram of cease and shoot, it is easier to differ. Spectrogram of cease and shoot

$\endgroup$
0
$\begingroup$

You cannot detect these two words as you are doing. The features you're using, i.e. the waveform, does not have the discriminability you're looking for. Speech recognition (the problem here) uses features that are much more cooked (MFCCs, etc)

I recommend using a popular open-source speech recognition packages like Sphinx3/4 (Java/C using HMMs) or Kaldi (C++/neural nets).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.