Real-time Single-channel Speech Enhancement with Recurrent Neural Networks

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=r6Ijqo5E3I4


Duration: 1:03:06
8,526 views
145

Single-channel speech enhancement using deep neural networks (DNNs) has shown promising progress in recent years. In this work, we explore several aspects of neural network training that impact the objective quality of enhanced speech in a real-time setting. In particular, we base all studies on a novel recurrent neural network that enhances full-band short-time speech spectra on a single-frame-in, single-frame-out basis, a framework that is adopted by most classical signal processing methods. We propose two novel learning objectives that allow separate control over expected speech distortion versus noise suppression. Moreover, we study the effect of feature normalization and sequence lengths on the objective quality of enhanced speech. Finally, we compare our method with state-of-the-art methods based on statistical signal processing and deep learning, respectively.

Slides: https://www.microsoft.com/en-us/research/uploads/prod/2019/09/Real-time-Single-channel-Speech-Enhancement-with-Recurrent-Neural-Networks-SLIDES.pdf

Learn more about the Audio and Acoustics Research Group: https://www.microsoft.com/en-us/research/group/audio-and-acoustics-research-group/



Other Videos By Microsoft Research


2019-09-19Recent Advances in Unsupervised Image-to-Image Translation
2019-09-19Modeling User Experience in Games: Lessons Learned
2019-09-18HCI, IR and the search for better search with Dr. Susan Dumais [Podcast]
2019-09-17Efficient and Perceptually Plausible 3-D Sound For Virtual Reality
2019-09-17Dashboard Mechanisms for Online Marketplaces
2019-09-13Fly-through in the AirSim simulation Team Explorer created for the DARPA SubT Challenge
2019-09-12Compacting the Uncompactable: The Mesh Compacting Memory Allocator
2019-09-12Privacy-Preserving Statistical Learning and Testing
2019-09-12Program Synthesis meets Notebooks
2019-09-11Inside the Microsoft AI Residency Program with Dr. Brian Broll [Podcast]
2019-09-10Real-time Single-channel Speech Enhancement with Recurrent Neural Networks
2019-09-06AI Institute Geometry of Deep Learning 2019 [Workshop] Day 2 | Session 1]
2019-09-05AI Institute "Geometry of Deep Learning" 2019 [Workshop] Day 3 | Session 1
2019-09-05AI Institute "Geometry of Deep Learning" 2019 [Workshop] Day 3 | Session 2
2019-09-05EzPC (Easy Secure Multi-party Computation)
2019-09-04Time discretization invariance in Machine Learning, applications to reinforcement learning...
2019-09-04Antennas for light and their applications in classical optics, Dr Rupert Oulton, Imperial College
2019-09-04Photonics for Computing: from Optical Interconnects to Neuromorphic Architectures
2019-09-04Inclusive design for all, or ICT4D and 4U! with Dr. Ed Cutrell [Podcast]
2019-09-03AI Institute "Geometry of Deep Learning" 2019 [Workshop] Day 2 | Session 4
2019-09-03AI Institute "Geometry of Deep Learning" 2019 [Workshop] Day 2 | Session 3


Tags:
Speech Enhancement
Neural Networks
deep neural networks
DNNs
Audio and Acoustics
enhanced speech
audio processing
speech quality
neural network training
Microsoft Research