Re-sizing sound signals

Hi All, I have sound signals of a database.each signal has different size. i want to do feature extraction. but due to different sizes of sound signals my feature matrix is giving me dimension mismatch.I want to know how to equalize length of those signals.

[Walter Roberson]

Are the samples at different frequencies and should be brought to a common frequency? Should the shortest represented time be used for matching? Should short samples be padded with 0's? Should the samples be considered to be periodic and repeated until they are all the same length?