Dear Readers
I am doing study of dimensional stones(marble/granite). I have
collected lot of signals using hammer and accelerometer setup from
NI. Now I am analyzing them. My target to find out out
frequency/energy concentration bands. I like to use wavelet packet
transforms and finally EMD on these decomposed signals.
I need to find out suitable mother wavelet for decomposition for
these type of signals. Does any body has already know about it?
Please let know.
I tried finding by max correlation of signal and constructed signal
using wavefun of matlab for all the wavelets and selecting highest
correlation. For every signal I get different mother wavelet. Does it
Ok?
Please suggest and comment to proceed further.
Thanks in Advance.
Mahabir Prasad.
Choosing Mother wavelet depends on many technical aspects starting from Actual acquired signal pattern in terms of frequencies and time information, noise and wanted signal etc.
Hi Doraisamy,
Yes, I am agree with you. I have vibration signals which are like transient signal and differs in duration (20 to 40 msec), available frequency components (500Hz to 6000Hz). I need to find dominate frequency bands in transient. Can you suggest probable mother wavelet to decompose.
Regards.
Mahabir,
It is true that you have a vibration signal against time, but this is the output of a hammer-accelerometer test, which means that the whole result is transient. Wavelets are good when you have continuous signal with time, and there are different transient moments that you want to analyze. but in your case there is only single transient output. I do not think that wavelet are the best option for hammer-accelo test. I think that FFT should give you the major frequencies. If you want to see how the frequencies change, then use Morlet, Shanon, or Mexican hat. If you think you have singularities, then use Daubichies,Symlets, or similar orthogonal wavelet of the low orders ~ Db4.
Best of luck with your research
Dear Tariq,
Thanks for reply.
I am trying look freq changes with time(distance) using Hilbert Huang Transform(HHT), while do so, results were not clear and there too much overlapping of imfs. While reading literature, there was suggestion-> Apply Wavelet decomposition and use HHT on these wavelets parts and add them finally. While doing mother wavelet is important. I stuck on it, tried to see literature and used them, still I am not sure that does it proper because most of literature talk about your suggested methods where as while selecting, coif5 find using corrcoef method.
Yes, I am looking for singularities, will try your suggested methods.
I am very new for wavelets being electrical engineering(power) student. Any lead will always welcome.
Regards
- Badges
- Science topic