i am an undergraduate working on a thesis for phoneme recognition using discrete wavelet transform (DWT), i was wondering if anyone could help me on what's the best method or algorithm to use for feature extraction of a signal. I will use this method to extract features so i can finish the templates i will use for my database.I already pre-processed the signals using dynamic time warping, then normalized them, then the next step will be feature extraction but i don't know what algorithm (which is not too hard) to use. I am trying standard deviation now and might consider other statistical-based method such as mean, variance, etc. if anyone can help me it would certainly be of big help or if you have an m-file of that algorithm it would surely help.thanks!!! |