Hi i am new to this group i am working on speech recognition for my last year project i have designed a gui in matlab for recognizing vowels and displaying them and used neural networks the problem is i have used filterbank coefficients(mel scale), lpc, and mfcc(probably as there are still some problems) to train the NN i dont know how to extract features from these co-eff i have decided on the co-eff they can be - formant locations, ratio of high to low energies, nasality, fricatives also, this is speaker independent(multi-speaker), so i am thinking of taking ratios of formant locations any help on the same would be welcome
Speech Recognition
Started by ●April 12, 2006
Reply by ●April 12, 20062006-04-12
hi, can you clarify your problem ? does the problem is in the choosing the way of NN input or the feature extraction (before entering to NN).
Reply by ●April 13, 20062006-04-13