A Novel Approach for Simultaneous Gender and Hindi Vowel Recognition Using a Multiple-input Multiple-output Co-active Neuro-fuzzy Inference System

Human beings can simultaneously recognize vowels in speech as well as gender of a speaker inspite of high variability. However, machines have not been able to simultaneously overcome both gender variability and vowel variability existing in speech due to gender. This paper uses a Multiple-Input Multiple-Output CoActive Neuro-Fuzzy Inference System to recognize both these patterns in speech simultaneously. The features used as input for the recognition is the pitch and the set of first three formant frequencies extracted from speech samples recorded from 70 Indian speakers, 33 male and 37 female. The individual recognition of either gender or vowel has been achieved at a rate of 68% and 95%, respectively, whereas the simultaneous recognition of both patterns has been attained upto 66% for the training set. Thus, this combined approach is a consolidated single-step novel approach which can replace the two-step method in automatic speech recognition systems where gender recognition is being used as the first step as part of hierarchical decision tree based vowel recognition. This can prove significant in enhancing the performance of an automated speech recognition system by eliminating an additional step.