Application of Deep Neural Network Algorithm in Speech Enhancement of Online English Learning Platform

INTRODUCTION: In the online English learning platform, noise interference makes people unable to hear content of teaching clearly, which leads a great reduction in efficiency learning. order improve voice quality speech enhancement method platform based on deep neural network is studied.OBJECTIVES: This paper proposes network-based for obtain more desirable results application optimization.METHODS: The optimized VMD (Variable Modal Decomposition) algorithm combined with Moth-flame optimization find optimal solution value decomposition mode number and penalty factor variational modal algorithm, then used filter information signal; Through learning, denoised signal taken as target achieve enhancement.RESULTS: research show that not only has significant denoising ability signal, but also after this used, PESQ perception evaluation greater than 4.0dB, spectral features are prominent, improved.CONCLUSION: experiments from three perspectives: denoising, spectrum information, usability confirmed.