The peak detection of the analog signal is very well known.
Now I will sample any analog signal, i'll convert it into a digital data.The resultant digital data peak has to be detected.
I know that for this traphizoidal filter is used. but details about this I have no idea.
You want to filter the signal too? Ok, prior to doing the peak detect, you can pass the signal through a FIR filter that has a trapezoidal impulse response. (Trapezoidal seems like a specialized application such as high-energy physics.) That's a straightforward FPGA processing stage - a few lines of HDL or a core pulled from the FPGA vendor's library.
yes trapezoidal filter is right. The signal here has no fixed baseline. ie refference may shift at any time.
So using trapezoidal filter is it possible to calculate the peak exactly?
What logic i can use in digital domain before the filter for the baseline correction?
One way of removing baseline shift is to measure its average value while the input is quiet. Then you digitize your interesting signal, and subtract the baseline value.
Why do you want to filter the signal? Filtering could distort the pulse shape and therefore its peak value. I realize that trapezoid filters have useful properties, but ... Maybe if you explain your application, then someone can give you more specific advice.