Utilizing Full Signal Reconstruction and Leveraging Perception for Deep Learning-Based Noisy Speech Enhancement

Deep learning has helped make speech processing applications, such as speech recognition, speech synthesis and speech translation, more prevalent in everyday life. It has also advanced the field of speech enhancement, where it has become the state-of-the-art approach to removing unwanted sounds. Successful speech enhancement can have profound impacts on how people communicate with each other and through electronic devices, so it is an area of research that must be addressed. Although several deep learning approaches have been developed, which focus on novel architectures and optimization strategies, there remains other aspects of the problem that have not been adequately investigated.

In this talk, we will discuss how human perception should be leveraged to further address problems within speech enhancement, and we will discuss how human perception can be predicted and used to improve noise reduction. Additionally, we will provide highlights of my work in complex-domain speech enhancement, which encouraged full signal reconstruction.