In this Medium article, I explore the use of SpeechBrain models to remove background noise from speech. I find that using these models, specifically the ResNet50-PASE and Conformer-CTC models, can effectively remove noise and improve speech quality. I also discuss the importance of data augmentation and the need for larger datasets to improve model performance. Overall, the results show that these models are a promising approach for noise reduction in speech signals.

Please read here.