Efficient Speech Signal Accent Recognition Algorithm based on Feature Fusion

Abstract

This study introduces a novel accent recognition algorithm using feature fusion. By extracting discriminative features from audio signals and applying evidence theory to feature information fusion, the algorithm effectively recognizes accents in speech signals. Our research examines popular speech signal processing models, including hidden Markov models, Gaussian mixture models, and methods based on deep and convolutional neural networks. Addressing the limitations of current methods, we propose a novel model to improve feature extraction accuracy and overcome the drawbacks of traditional methods. An original feature fusion algorithm flexibly combines multiple speech signal features, exploiting their complementary advantages, optimizing the fusion process, and ensuring a more powerful and comprehensive representation for subsequent analysis. Experimental validation demonstrates the significant superiority of the algorithm, marking a decisive advance in accent recognition. The innovative approach positively impacts the broader field of speech signal processing, reaching new heights in accent recognition. The results show significant advantages in accuracy and efficiency over traditional methods, confirming the practical potential of the algorithm and providing valuable insights for future accent recognition research.