machine learning - Are MFCC features required for speech recognition -

i'm developing speech recognition project , i'm trying select meaningful features. of relevant papers suggest using 0 crossing rates, f0, , mfcc features therefore i'm using those. question is, training sample duration of 00:03 has 268 features. considering i'm doing multi class classification project 50+ samples per class training including mfcc features may suffer project curse of dimensionality or 'reduce importance' of other features. question is, should include mfcc features if not can suggest alternative?

you should not use f0 , 0 crossing, unstable. can increase training data , use mfccs, have representation capabilitites. remember mean-normalize them.