Grant: Alliance Grants

The advent of high-speed, ultra-low latency 5G wireless networks will foster the emergence of new, varied, and more natural forms of interactions providing the "being there, here" kind of experience. However, existing speech and audio coding standards do not yet enable the required ultra-low latency while maintaining a high subjective audio quality, a low bit rate and a manageable complexity. Furthermore, to constitute an appealing proposal for application developers and service providers, next generation speech and audio coding standards should provide additional benefits such as the ability to manipulate the encoded information for various purposes: speaker or audio object identification, audio content transformation, and even mapping to and from other modalities such as text, image, and video. The main goal of this project is therefore to develop a new generation of very low latency speech and audio coding tools and solutions for immersive interactions. This project will produce innovations in two main axes: 1) efficient ultra-low latency speech and audio compression and 2) high level feature analysis and synthesis. A large part of the project will expand on current machine learning tools and neural network architectures, including bio-inspired signal processing and spiking neural networks. A baseline ultra-low latency speech and audio codec will also be studied and developed by exploring the application of predictive and adaptive quantization to the statistics of micro-frames a few audio samples in duration. This project is important to Canada and its economy because of the huge contribution which can be made to several areas including live and streamed presentations of sports and cultural events, mobile communications, gaming, virtual and augmented reality. As an industrial partner, VoiceAge has already demonstrated the huge impact it can have by licencing technologies positioned in many worldwide standards in speech and audio compression. Ultimately, this project will allow Canada to remain a world leader in the field of speech and audio compression for a wider range of applications.