Exploring the Influence of Noise on Voice Recognition Systems: A Case Study of Supervised Learning Algorithms

Abstract

Speech recognition systems have become increasingly prevalent in various applications, ranging from virtual assistants to voice-controlled devices and dictation software. The desire to design human-computer interfaces that are more accessible, intuitive, and efficient cannot be overemphasized. Such advancements hold significant potential to enhance accessibility, productivity, safety, and user experience across various domains and industries. Despite significant advancements in speech recognition technology, several challenges persist. One primary challenge is achieving high accuracy and robustness across diverse speaking styles, accents, and environmental conditions. Background noise, variations in pronunciation, and overlapping speech can introduce errors and degrade performance, especially in real-world scenarios. This study explores the pursuit of accurate and lightweight algorithms in biometric applications, focusing on voice recognition. Using a dataset featuring renowned leaders' voices, the research compares K-Nearest Neighbors (KNN) and Artificial Neural Network (ANN) techniques. Employing Mel Frequency Cepstral Coefficients (MFCC) and considering noisy and noiseless environments, the study reveals that ANN outperforms KNN. In noisy conditions, ANN achieves 62.4% accuracy, while KNN reaches 33.4%. In noiseless settings, ANN's accuracy rises to 95%, surpassing KNN's 88%. Assessment metrics like False Acceptance Rate, False Rejection Rate, F1 Score, Recall, Precision, and Receiver Operator Characteristics Curve are analyzed. The study emphasizes the detrimental impact of noise on recognition accuracy and underscores ANN's consistent superiority over KNN. Despite challenges posed by noisy environments, the research highlights the potential benefits of these approaches, emphasizing ANN's superior performance across scenarios. This research showcases the significance of accurate biometric systems, emphasizing ANN's advantages in enhancing usability and precision, especially in real-world noisy conditions.