Detect fundamental Frequency

>On Jan 7, 10:46 am, robert bristow-johnson <r...@audioimagination.com>
>wrote:
>
>> sounds like what you need is what we call in the audio/music DSP
>> world, a "pitch detection algorithm".  like AMDF or ASDF or
>> autocorrelation.
>
>Siimilarly, in the voice processing world there is, for example:
>
>RAPT (Robust Algorithm for Pitch Tracking), David Talkin, in "Speech
>Coding and Synthesis", edited by Kleijn and Paliwal.  You'll find an
>implementation in the Matlab VoiceBox Toolkit, available from  "http://
>www.ee.ic.ac.uk/hp/staff/dmb/voicebox/voicebox.html".  Finding a copy
>of the original article is extremely difficult, however.
>
>HPS (Harmonic Product Spectrum),  M. R. Schroeder, "Period Histogram
>and Product Spectrum:  New Methods for Fundamental-Frequency
>Measurement", The Journal of the Acoustical Society of America, Volume
>43 Number 4, 1968.
>
>Greg
>


Nice article.... This is what I was looking for... many thanks for
everybody

fredp wrote:
>>fredp wrote:
>>
>>
>>>Hello All,
>>>
>>>I am trying to detect the fundamental frequency of a non-sinusoidal
>>>waveform. I do FFT on the sampled data and I get a bunch of harmonics.
> 
> How
> 
>>>should I go about to detect the fundamental frequency between the
> 
> harmonic
> 
>>>signals?
>>>
>>>Thank you
>>>
>>>
>>
>>Look up definition of "fundamental" and "harmonic".
>>IOW you didn't ask the question you meant to ask ;)
>>
>>
> 
> 
> Thanks Richard,
> Well let me try to clear things up. I have got a square waveform. When I
> put it in the frequency domain I get the fundamental frequency and
> harmonics ok? Lets say the partial harmonics are greater than the
> fundamental component, So the peak fft bin is not related to the
> fundamental freq but to the partial one. For example, if a waveform has
> three components at 2,4 and 6 Hz and assume that the 4 hz component is
> stronger than 2 hz....
> 
> Thanks

I think you are still you are missing something in visualizing your problem.

Informative exercises.
Calculate Fourier _series_ of a square wave.
With whatever fft package you are using, calculate fft of an *INTEGER* 
number of full cycles of an ideal square wave. Plot it.
Repeat for a non-integer number of cycles. Plot it.
See also:
http://en.wikipedia.org/wiki/Fourier_series (and links)

>fredp wrote:
>>>fredp wrote:
>>>
>>>
>>>>Hello All,
>>>>
>>>>I am trying to detect the fundamental frequency of a non-sinusoidal
>>>>waveform. I do FFT on the sampled data and I get a bunch of
harmonics.
>> 
>> How
>> 
>>>>should I go about to detect the fundamental frequency between the
>> 
>> harmonic
>> 
>>>>signals?
>>>>
>>>>Thank you
>>>>
>>>>
>>>
>>>Look up definition of "fundamental" and "harmonic".
>>>IOW you didn't ask the question you meant to ask ;)
>>>
>>>
>> 
>> 
>> Thanks Richard,
>> Well let me try to clear things up. I have got a square waveform. When
I
>> put it in the frequency domain I get the fundamental frequency and
>> harmonics ok? Lets say the partial harmonics are greater than the
>> fundamental component, So the peak fft bin is not related to the
>> fundamental freq but to the partial one. For example, if a waveform
has
>> three components at 2,4 and 6 Hz and assume that the 4 hz component is
>> stronger than 2 hz....
>> 
>> Thanks
>
>I think you are still you are missing something in visualizing your
problem.
>
>Informative exercises.
>Calculate Fourier _series_ of a square wave.
>With whatever fft package you are using, calculate fft of an *INTEGER* 
>number of full cycles of an ideal square wave. Plot it.
>Repeat for a non-integer number of cycles. Plot it.
>See also:
>http://en.wikipedia.org/wiki/Fourier_series (and links)
>


After searching around and reading a few articles here and there, I found
about cepstrum analysis to be of great use in finding the fundamental
frequency. Is it reliable? How is it implemented?

On Jan 7, 12:01&#4294967295;pm, "fredp" <fred...@gmail.com> wrote:

> After searching around and reading a few articles here and there, I
> found about cepstrum analysis to be of great use in finding the
> fundamental frequency. Is it reliable? How is it implemented?

For my purposes in voice processing, I found it to be useless.  YMMV.

Alan V. Oppenheim and Ronald W. Shafer, Digital Signal Processing
(Englewood Cliffs, New Jersey: PRENTISS-HALL, Inc., 1975), 518-520.

Greg

Hi,

First is auto-correlate the signal with itself. This will
help in filtering of any noise. now the periodicity can be
found by means of zero crossing detectors.

Bharat Pathak
www.arithos.com





>Hello All,
>
>I am trying to detect the fundamental frequency of a non-sinusoidal
>waveform. I do FFT on the sampled data and I get a bunch of harmonics.
How
>should I go about to detect the fundamental frequency between the
harmonic
>signals?
>
>Thank you
>
>
>

Greg Berchin wrote:
> 
> On Jan 7, 9:53 am, "fredp" <fred...@gmail.com> wrote:
> > For example, if a waveform has
> > three components at 2,4 and 6 Hz and assume that the 4 hz component
> > is stronger than 2 hz....
> 
> It's not at all uncommon for some harmonics, or even the fundamental,
> to be weak or missing entirely.  In such cases you can often infer the
> fundamental frequency from the differences between the frequencies of
> the harmonics that are present.

Right. The inference for the frequencies given by the OP (2,4 and 6) is
that the fundamental freq. is 2 and therefore not at all missing as the OP
seems to claim. That is, he only thinks its missing. If the frequency was
2,4 and 5 the wave pattern would repeat every 1 Hz. 

-jim


> 
> Greg