implementation of Cepstrum algorithm

[madgreek]

hello guys.this is my first ever post in any kind of a discussion room.i read some of your posts and some of you out there sure know your stuff 🙂

well basically this is what im dealing.im trying to time-pitch scale a speech signal (or maybe a monophonic one but mostly speech) using the TD-psola algo.i know u have to use firstly a pitch detection algo and in my case i thought to use the Cepstrum one. ifft(log (abs)fft(x)).ideally speaking the outcome of the cepstrum would be a clear peak denoting the fundamental frequency of the signal , right?well it doesn't.i dont know what im doing wrong.my best guess would be somewhere the log function is being used but not sure.does anybody can give me any tips?somewhere he thinks i might (probably am) be wrong?any help would really be appreciated

kind regards

alex

[unclebump]

See this NI manual.

http://www.ni.com/pdf/manuals/371532a.pdf 

[madgreek]

hello again and thank you for your quick reply.it did help but not to the point that i was hoping for.do you think something is wrong with my program?can you give me any tips on doing that?

any help is wellcome

regards

alex

[madgreek]

i have acquired a 10 sec of my speech in labview.i want to estimate the pitch period of it using the cepstrum algo IFFT(LOG(FFT(X))) in order to use it later on in the TD-PSOLA. but something is not working right.i found another version of the cepstrum using fft in place of the ifft.whats the difference?which one should i use?should the output of the cepstrum be really a clear peak or it needs further processing?does anybody know what i should do?i'll appreciate any help guys

kind regards

alex

p.s i have attached my program in the first thread