Image Feature Extraction for Signature Recognition

I am currently doing a Signature Recognition System for my academic achievement. I would like to use the image features as the input data for my neural network. But i don't know which features are relevant for the recognition. Anyone can give me some suggestion? Thank you.

Answer 1

[Walter Roberson]

We do not know either. What we would do to find out is to search for papers and books that discuss the topic. Try http://scholar.google.com . The choice of feature is not a question about MATLAB, but once you have picked one particular one, we can help you if you encounter problems with the implementation.

[Vishnu Rajen]

Thanks for your reply, I did went through some papers to find out. So my choices are kurtosis, skewness, solidity, orientation, entropy, mean, standard deviation, eccentricity, and euler number. However, these are from the papers i found but they don't explain why i need these features? I was wondering maybe i can get some briefly explanation here?

[Walter Roberson]

Keep reading. Look at the definitions of each of them and ask yourself "How would this help me distinguish between letters? What are some letters for which the values would be different?"