AI Speech talk from Facebook ML Engineer: Speech Synthesis after WaveNet
Event Information
About this event
Date: April 27, 2021
Time: 6PM – 7:30PM
IEEE SCV Young Professionals, Computer Society and Computational Intelligence Society invite you to a seminar on the latest research and implementation on AI Speech Architectures including WaveNet and its successors.
Abstract: In 2016, Deepmind proposed WaveNet which became the state of the art for realistic speech synthesis using an autoregressive deep neural network. After that, many new ideas based on autoregressive modeling are proposed, such as Parallel WaveNet, WaveRNN, etc.
In recent years, there is another research direction that begins to draw more and more attention. That is using Generative Adversarial Network(GAN) for audio generation. In today's talk, I will give an overview of some noteworthy architectures since WaveNet and some promising GAN-based models from recent years.
Bio: Jason (Jilong) Wu is currently ML Engineer@Facebook AI Applied Research(FAIRAR). Focused on AI-driven speech synthesis, self-supervised learning and edge device model optimization. Previously keyword spotting team@Google Speech.
Cornell B.S. and Meng, Electrical and Computer Engineering, specialized in ML and embedded systems.
