# Finding peaks on a curve

• May 9th 2011, 02:29 AM
malaguena
Finding peaks on a curve
Hello,

I have a variable number of points (not more than 10) - which are probability values. Amongst these points some of them refer to the peaks on the curve if the points are fit on a curve. I want to find the points that represent the peaks on the curve. Is there a simple way to find the peaks? Do I need to do any curve fitting first to spot the peaks?

Thanks..
• May 9th 2011, 02:57 AM
CaptainBlack
Quote:

Originally Posted by malaguena
Hello,

I have a variable number of points (not more than 10) - which are probability values. Amongst these points some of them refer to the peaks on the curve if the points are fit on a curve. I want to find the points that represent the peaks on the curve. Is there a simple way to find the peaks? Do I need to do any curve fitting first to spot the peaks?

Thanks..

I'm afraid we would need to know a lot more about the data than you have posted to make any sensible suggestions.

CB
• May 9th 2011, 03:27 AM
malaguena
Quote:

Originally Posted by CaptainBlack
I'm afraid we would need to know a lot more about the data than you have posted to make any sensible suggestions.

CB

The problem is to find the split points in a word (morphological segmentation). For each split point, my model assigns a probability. The word is to be split at the points where the probabilities are high. For example, in the word "walkings", there are 8 possible split points that each has got a probability. The point after "s", and the point after "ing" are expected to be comparably higher than others.

To give an idea an array of logarithms of these probabilities could be:

1093180.2921060047 -1099512.0227943386 -1095977.4780250955 **** -1101739.5058371008 -1103636.1269468896 -1103636.1269468896 -1103636.1269468896 -1103636.1269468896 -1103636.1269468896 - 1096910.8541777392 **** -1103636.1269468896 -1103636.1269468896

Here the split points that need to be found are marked with ****.