Auto Threshold detector

[scartch]

i am doing an ECG project. MY program is unable to solve baseline wandering problem, so if i have different type of data and the threshold will be different and i need to set threshold myself. Is there any way to solve the problem instead of seting threshold by myself. My threshold need to detected the most numbers of peak in order to perform pulse rate calculation.

i am using ver 8.2 student

[TiTou]

Hello,

I suggest to calculate the point by point derivative of your signal and detect when it crosses zero from positive to negative.

Hope this idea helps

PS : you posted your question into the "machine vision" section the forum... I think it would make more sense to post it in the LabVIEW section


Message Edited by TiTou on 01-22-2008 11:47 AM

[griffindore]

Hi scartch,

Although TiTou's suggestion of using the derivative will definitely work to find all the peaks, I do not believe it will work with your particular application. The purpose of the Threshold Peak Detector is to eliminate any false peaks that may be introduced by noise. For instance, the data you provided would count peaks for the transitions between .003 and .005 and between .005 and .007. These peaks will be counted due to noise in your signal, but the Threshold Peak Detector will filter it out with the use of the threshold and width inputs.

This question is very similar to this thread. I suggest using the original thread since it is on the appropriate board. As Jennifer alluded to in the original thread, I suggest taking subsets of your data, finding the average, and normalizing the data (by dividing the average). Then you can use the same threshold and width for all of your data. Appropriate values for threshold and width will depend on the data. It is a very common method to use a moving average filter (described above) to correct the baseline wandering problem. I hope this helps!