ДОСЛІДЖЕННЯ КАПСУЛЬНИХ НЕЙРОМЕРЕЖ ДЛЯ РОЗПІЗНАВАННЯ ПРИРОДНЬОГО МОВЛЕННЯ

This thesis focuses on evaluating neural network-based models for speech recognition under noisy conditions. It examines the effectiveness of the capsule neural network across various types and intensities of noise. By gathering and preprocessing a dataset of noisy speech recordings, the research compares these models based on their recognition accuracy, resilience to different noise types and intensities, and computational efficiency.