-1

Source Code

library("tuneR")
library("seewave")

#0: Acquisition of sample sound
snd_smpl = readWave(paste("~/Music/sample/1980s-Casio-Celesta-C5.wav"), 
                                    from = 0, to = 1, units = "seconds")
dur_smpl = duration(snd_smpl)
len_smpl = length(snd_smpl)

#1 : Pre-Processing Stage

#1.1 : Application of Hanning Window
n = 1:len_smpl
han_win = 0.5-0.5*cos(2*pi*n/(len_smpl-1))
wind_sig = han_win*snd_smpl@left

#2.1 : Auto-Correlation Calculation
rev_wind_sig = rev(wind_sig)    #Reversing the windowed signal

acorr_1 = convolve(wind_sig, rev_wind_sig, type = "open")
# Obtaining the 2nd half of the correlation, to simplify calculation
n = 2*len_smpl-1 
acorr_2 = (1/len_smpl)*acorr_1[len_smpl:n]

#2.2 : Note Calculation
min_index = which.min(acorr_2)
print(min_index)
fs = 44100              
fo = fs/min_index #To obtain fundamental frequency

print(fo)
print(notenames(noteFromFF(fo)))

Output

> print(min_index)
[1] 37
> fs = 44100                
> fo = fs/min_index 
> print(fo)
[1] 1191.892
> print(notenames(noteFromFF(fo)))
[1] "d'''"

The entire calculation is performed in the Time Domain. I'm currently using autocorrelation as a base to understand more about Pitch Detection & Analysis. I've tried to analyse the sample with 'Audacity' and the result is 'C5'. Hence, I'm wondering where actually the issue is. Can you all help me find it?

Also, there are a few but important doubts:

  1. How small should actually my analysis window be (20ms, 1s,..)?
  2. Will reinforcement of the Autocorrelation Algorithm with AMDF and other similar algorithms make this Pitch Detection module more robust?
Argon
  • 101
  • i have a couple of answers about how i would suggest approaching pitch detection in the time domain. 1, 2. there is a relationship between autocorrelation (ACF) and the average squared* difference function* (ASDF) as an alternative to average magnitude difference function (AMDF). also, if you do ACF of a finite length with convolution, you get a tapering effect. – robert bristow-johnson Oct 11 '18 at 06:19

1 Answers1

0

I'm not familiar with the language you're using, but some time ago I made a really simple pitch calculator for voice processing using autocorrelation of the signal, as you may know, if a signal is periodic it's autocorrelation will be periodic and both will have the same period, and in order to identify it's period you have to find peaks in the autocorrelation of the signal, for example in this figure we have a voiced signal:

enter image description here

Don't pay attention to the horizontal lines, and the autocorrelation of this signal looks like this:

enter image description here

In the latest image you can see the periodicity of the autocorrelation, and in order to identify the period of the signal I used a treshold of $0.3 r_c[0]$ (where r_c[0] is the maximum value of the autocorrelation), and find the first peak that is higher than the threshold inside a "known" region (in my case it was between the samples 20 and 200 because of the nature of the signal).

The size of the windows will depend on your signal, in case of voice processing the size is usually between 10~30 ms.

Hkxs
  • 114
  • 5