A Comparative Study of Machine Learning Algorithms for Speaker Identification Using MFCC and Pitch

Abstract

The primary objective of this project " A Comparative Study of Machine Learning Algorithms for Speaker Identification Using MFCC and Pitch" is to create a reliable and effective system for accurately recognizing speakers through audio input. This project utilizes advanced machine learning methods and audio feature extraction techniques, with a primary focus on Mel-Frequency Cepstral Coefficients (MFCC) and pitch features, to effectively categorize speakers. During the training process, the system obtains characteristics from clear audio sets and utilizes them to train various machine learning models like Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Gradient Boosting. These models are assessed using important performance criteria such as accuracy, precision, recall, and F1-score. During the testing stage, the trained models are assessed on audio data that has not been seen before in order to evaluate their strength and ability to apply to new situations. This app showcases how feature extraction methods and machine learning algorithms can be combined effectively to improve the accuracy of speaker recognition. Python is used for the implementation of the system, making use of libraries like Librosa to extract features and Scikit-learn for training and evaluating models. The findings show that this method has promise for practical use in security systems, call centers, and smart assistants. This project offers a solid foundation for speaker recognition and also opens doors for more studies in audio-focused machine learning technologies.