Gender Voice Recognition Using Random Forest Recursive Feature Elimination with Gradient Boosting Machines

Speech emotion recognition is a difficult task in the field of affective computing because emotions in speech heavily depend on a variety of factors such as feeling, thought, behaviour, mood, temperament, personality and disposition that are hard to model. Emotion plays a significant role in decision making and it influences human perception, learning, behaviour and relationships between individuals. Gender voice is a contributing factor in boosting the accuracy of emotion recognition systems using speech signals. In this paper, we propose a gender voice recognition method which makes use of feature selection through the Random Forest Recursive Feature Elimination (RF-RFE) algorithm with Gradient Boosting Machines (GBMs) algorithm for gender classification. The training and testing data were obtained from a public gender voice dataset. The GBMs algorithm was later evaluated against the feed forward neural network and extreme machine learning algorithms. The classification accuracy of the GBMs improved after applying the RF-RFE to the dataset. Experimental results indicate that GBMs outperformed all the comparative algorithms in classification accuracy and proved to be a suitable candidate for gender voice recognition.