speaker
Oxford University logo
Phonetics Laboratory
Faculty of Linguistics, Philology, and Phonetics

An example of voicing detection

The voicing signal produced by using voicing.c to analyse joe.dat is shown in figure 5.1. If joe_voice.dat is used to mask joe_cepstral_f0.dat ( figure 4.11 ), by multiplying each sample of joe_cepstral_f0.dat by the corresponding sample of joe_voice.dat (using the program multiply.c ), the f0 signal in figure 5.2 is obtained.

Figure 5.1. Voicing of part of joe.dat

Voicing of part of joe.dat

Sample number (× 104)

 

Figure 5.2. Cepstral f0 of part of joe.dat, masked by the voicing estimate

Cepstral f0 of part of joe.dat, masked by the voicing estimate

Sample number (× 104)


However, a different method of f0 estimation may give better results ...

Figure 5.3. f0 of part of joe.dat, estimated by the autocorrelation  method, masked by the voicing estimate

f0 estimated by the autocorrelation method

Sample number (× 104)



Next: The autocorrelation  method of  f0 estimation