Recognition of English spoken stressed syllables based on natural language processing and endpoint detection algorithm

Abstract

The traditional stress evaluation method cannot describe the chaotic characteristics of speech signals well or cannot fully approximate the complex nonlinear relationship between features, so that the exact location of stress cannot be accurately determined. In order to improve the recognition of spoken English stressed syllables, this study established a word stress recognition model. The model can accurately recognize the stressed syllables in the words and calculate the fundamental frequency change trajectory of the language tones using the fundamental frequency scaling function. According to this trajectory and using the time domain fundamental frequency synchronization superposition algorithm to modify the fundamental frequency parameters in the DIVA model motion instruction, the spoken language learner can master the rhythm in the spoken language. In addition, this study sets up experiments to study the effects of the model. The results show that the stress recognition model established in this paper has good reliability and stability.