A Novel Approach for Multi-pitch Detection with Gender Recognition

Abstract

An audio signal can be estimated for its frequency ad saliencies of pitches sounds using multiple pitch estimation for a short-time frame. The focus of this paper is to solve the difficulty for predicting the fundamental frequencies of a signal comprising various harmonically related sinusoidal signals. The framework proposed in this paper can be used to detect single as well as multi-F0 along with gender identification. The performance of the approach is evaluated on 135 mixture audio signals which compromise of human speech of different genders. The successful evaluation was performed for a number of human in the complex mixture speech signal and their respective genders. It is based on the iterative approach in which frame-by-frame analysis is carried out for multi-F0 detection. The algorithm effectively provides optimal results in classifying gender identification and person calculation.