How would you find the pitch period of a sound signal?

[madgreek]

Hello Mike

Thank you for your help.

Basically what you doing in your vi is taking different size FFTs thoughout the signal and by looking at the 3d graph you can fine-tune the sizes, am i right? I guess there are some differences from your version Labview to mine and some settings didnt load correctly thus i was not able to see the 3d graph.

Alex

[DuffmanSays]

Alex,

In the VI, I am taking the same size FFT throughout a single run of the program. Again, I have chosen this value to capture at least one period of the lowest frequency signal. This is so that for any window I examine, this frequency component will be visible. Thus I am not programmatically determining the window size to use.

For any FFT, we can just see a snapshot of the frequencies and how prevalent they are in the signal. In the example VI, each window takes several FFT snapshots that are then layered onto the 3D graph. Thus by looking at the graph I can see when any frequency shows up and disappears from the original signal.

The example shows this idea, so we may want to consider why it is that you cannot successfully run the VI. When you say that “some settings did not load,” what behavior is it that you see when running the VI?

Mike D.

[madgreek]

Hello Mike. How are you?

When i tried to open your vi i got the following message " Error loading control. A newer version is needed. The default settings for the control will be used" and although i am able to run the vi with no broken wires at all, the 3-D graph is not shown in the front diagram. I even replaced it but still the same thing. I guess it doesnt like me much   or im doing something wrong here.

Alex

[DuffmanSays]

Hello Alex,

I believe that this message is related to versions of ActiveX that may be installed on your machine for viewing the 3D graphics. In the past, I have seen reinstallation solve this problem to allow the 3D rendering capabilities to work correctly. If you should decide that you like the method that I have proposed, we can work toward getting this resolved. However, to simplify things, you will see the resulting 3D image attached below. As I have mentioned, this is composed by effectively stacking each FFT slice to form the entire curve. In this way, I can see what frequencies are most prevalent for any given time range. While you may not necessarily choose to plot these curves in 3D, this visualization may help you derive an algorithm to extract this information using the method in the example. Please post back if this helps or if I can further clarify. Thanks!

Mike D.

Message Edited by Duffman Says: on 04-24-2007 09:54 PM

[madgreek]

Hello Mike

That is a way cool way to visulize the whole approach. I have never used the 3-D graph (i am very novice when it comes to labview as you can see) but i really liked it. It would help alot to get a better idea about the whole approach. I will try and rebuild your code and get my own results, i hope it will help. If not i will see what i can find about the 3-d active-x updates for my computer (although is a brand new one, 4-5 months old, and i do all its updates frequently, i dont know why it doesnt support this.)

Thank you again Mike for all your help. I will keep you posted

[DuffmanSays]

I am glad that you like the idea and this (decidedly cool) LabVIEW feature. As I mentioned previously, the LabVIEW installation should include all the necessary ActiveX controls for 3D graphing. If any of these components were modified since the install, a reinstall might work to get the capability back for you. Anyway, post back if you're not able to get this working correctly. Thanks.

Mike D.
National Instruments
Applications Engineer

[johnsold]

Alex,

Have you looked at the WVD Spectrogram VI or the STFT Spectrogram VI? WVD is the Wigner-Ville Distribution and STFT is Short-Time Fourier Transform. Both are designed to show the time variations of spectral content. The WVD often has some spurious components but since you are trying to identify the fundamental, it should be relatively easy to avoid them. If nothing else, it will give you a visualization tool to compare to your other algorithms.

Another approach might be to use the Point by Point FFT functions with the sample set length set to the smallest chunkc size that is meaningful for your data.

Lynn