I'm working with 16-bit WAV files. As far as I understand, amplitude is measured at a sampling rate of 44100 samples/second (i.e. Hertz), with each sample being represented as binary data through quantization.

When I retrieve frequency statistics from a WAV file, Praat gives me quantile information i.e. it is approximating the distribution of the quantized binary data. How exactly is it doing this i.e. what algorithm does it use? In addition, can I retrieve the underlying distribution to perform statistical analysis on the raw data rather than its approximation?

Thank you.

  • Your sampling rate means that 44100 digital values are used to describe each second of your signal. So it is important your temporal accuracy. Your sampling resolution means that 2^16 values are used to describe all the amplitude. Binary data is smth else. – amegnunsen Oct 31 '19 at 11:30
  • Inside Praat, there is a help that can give you informations about quantile. With any computer programming (python, R, ...) you can carry out any statistical analysis on audio data. – amegnunsen Oct 31 '19 at 11:46
  • If you want to see the temporal distribution of frequencies, try a spectrogram. 16bit@44kHz is lossless for most intents and purposes. The frequency information is retrieved by transformation (Fourier) from the temporal to the frequency domain, averaged over short time slices around each point. The specifics are not linguistic – vectory Oct 31 '19 at 17:47

Your Answer

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

Browse other questions tagged or ask your own question.