i have achieved translation and rotation independence thru karhunen loeve algorithm now if i want it to be expansion independent wat do i do?

hi,
I have a problem..what is the algorithm to make a programme for the instantaneous autocorrelation function for the single phase voltage waveform measured at particular location i.e. nonstationary signal.
Actually i have data in (1283*2) matrix
1st column represent the time at a difference of 7.8e-5sec.
and second column represent the voltage magnitude at that particular instant.

Hai friends,
This is my problem:

i)Determine the mean and autocorrelation function for the following random process
X(t)=v(t)+3v(t-τ) where τ is a constant and v(t) is a stationary random process of independentrandom variable with mean μ and variance σ2

ii) Is X(t) wide sense stationary?

Problem:
Let:
X(t) and Y(t) be independent, WSS random process with zero mean and the same covarinace function Cx(t).
If Z(t)= aX(t) + bY(t), then:

a)Determine whether Z(t) is also WSS.

b)Determine the probability density function of Z(t), if X(t) and Y(t) are Gaussian random process.
Thanks.

Simple and single question:
I am doing my self a WEB page, when I start to do image transfers, my computer immediately start to slow down or finally stop for a while.
I am using a laptop COMPAQ presario 2000, the Device Instance Id is: ACPI\PNP02004&BF7C77B&0.
Is these device responsible for the delays?

can i ask a question
why the autocorrelation function must be a nonnegative function?
please help me, thnx so much!

Need help on figuring out the of the known problems with linear regression
Autocorrelation, multicollinearity and heteroskedacticity. Also what method to correct?

Thanks

Hi,
I am unable to grasp a very simple concept.
Imagine I have a data set which is 1 hour long and consists of delta pulses of constant height occuring randomly with the maximum occurence rate of 1 million pulses per second.

1. Either I treat it as one single data set and perform the time average. Let me call it and the units are counts per second.

2. I can also imagine to break the data set into smaller subsets each of 1 second interval i.e., I will have 3600 data subsets which now I call ensemble. Each data subset will have variable number of pulses. Now let me take a mean of number of pulses falling in each subinterval and call it and this will have the same unit i.e., counts per subset which in our case is 1 sec.

Apparently the two are not equal. I found this when I was calculating autocorrelation function (or coherence function as defined by physicsts - apparently it is not the same as statistical normalized correlation function). I tried method 1 and then tried method 2 for a case where I generated correlated data. Books tell me that second order coherence function for a correlated data should be greater than one. I can never see that happening when I perform operation 2. It is only when I do operation 1, that I see some hopes of the coefficient to be greater than one. And this is because - although not all 1 million photons occur in every 1 second interval, the fact that it may occur is taken into account when you do time averaging. In doing mean this fact is not taken into account.

Does it mean that when one calculates mean on data set, it is not the true ensemble average but good enough for certain purposes?

Please help me with my confusion.
I will appreciate it.
Thanks

Hi,
I'm troubled by the notation used in Example 1 on the side "Autocorrelation of Random Processes". The very last step reads as follows:
--Or in a more concise way, we can represent the results as--
Then the autocorrelation of a Gaussian random process becomes the variance times the delta function dependent on 'm'. My question is:

Does variance times delta function dependent on 'm' really describe the same relationship as the previous equation where two cases are considered?

I'm asking that question because for 'm'=0 the autocorrelation describes the power. However, the Fourier transform of the autocorrelation function describes the power density spectrum. Now, to calculate the power, the Equation with the different cases comes in handy, because to evaluate the power means simply to look at the case for 'm'=0, i.e. the power for white Gaussian noise is the variance. However, if I want to calculate the power density spectrum I need to evaluate the Fourier transform of the autocorrelation. In this case the form variance times delta function is preferable, because this yields the expected result. The power density spectrum of white Gaussian noise is a straight horizontal line at the height equal to the variance.

Thanks,
Oliver Faust

Hi!
I am working on the following exercise and i would like some help.
I have a sin-wave and i mix it with with noise.Therefore i am trying to find the minimum required SNR level, in order to achieve the same sin-wave as far as frequency is concerned. It is obvious that the SNR level must be above zero and therefore the sin-wave (for the frequency i use which is 0.37) must have at least width around 2 (2*sin(2*pi*0.37*n).
Is that enough or it needs more?
Thanx in advance!