Automatic speaker recognition is the most popular and challenging problem in the area of AI & ML. It is used in the field of recognizing the human voice and it is mainly used for user authentication for safety majors and finding a particular human from a lot of speakers. It is difficult to work to put in a machine the dissimilarity persons’ voices mainly with a variety of audio samples like accents, genders, language, etc. This paper includes the deep learning approach for setting up CNN and the neural networks, which were put on multiple takeout characteristic from audio samples, and is trained with various spectrograms. Transfer learning approaches are included to obtain a proper output utilizing a specific dataset i.e Kannada Kali. The proposed model gave an accuracy of up to 70%.
₹10000 (INR)
NON IEEE-2022