Improving Deep Speech Denoising by Noisy2Noisy Signal Mapping
This addresses the problem of speech denoising in real-world audio environments for applications like communication or audio processing, by eliminating the need for clean training data, which is an incremental improvement over existing deep learning methods.
The paper tackles speech denoising without requiring clean speech signals for training by using a self-supervised method that maps between two noisy versions of the same signal, and shows superiority over conventional supervised approaches in experiments using four performance metrics and field-testing.
Existing deep learning-based speech denoising approaches require clean speech signals to be available for training. This paper presents a deep learning-based approach to improve speech denoising in real-world audio environments by not requiring the availability of clean speech signals in a self-supervised manner. A fully convolutional neural network is trained by using two noisy realizations of the same speech signal, one used as the input and the other as the output of the network. Extensive experimentations are conducted to show the superiority of the developed deep speech denoising approach over the conventional supervised deep speech denoising approach based on four commonly used performance metrics and also based on actual field-testing outcomes.