A rough set theory and deep learning-based predictive system for gender recognition using audio speech

Speech is one of the most delicate medium through which gender speakers can easily be identified. Though related research has shown very good progress in machine learning, but recently, deep learning imparted a area to explore deficiency discrimination using traditional techniques. In techniques, speech features are automatically generated by reinforcement from raw data have more discriminating power than human-generated features. But some practical situations like recognition, it observed that combination both types sometimes provides comparatively better performance. proposed work, we initially extracted and selected informative precise acoustic relevant recognition entropy-based information theory Rough Set Theory (RST). Next, audio signals directly fed into neural network model consisting Convolution Neural Network (CNN) Gated Recurrent Unit (GRUN) for extracting useful recognition. The RST selects features, CNN extracts locally encoded important GRUN reduces vanishing gradient exploding problems. Finally, hybrid system developed combining feature vectors. been tested with five bench mark simulated dataset evaluate its performance, combined vector effective specially when transgender considered as type together male female.