Systems and methods for speech transcription

US10540957B2 · kind B2 · utility

Assignee

• BAIDU USA LLC · US

Inventors

• Awni Hannun · Los Altos, US
• Carl Case · San Francisco, US
• Jared Casper · Sunnyvale, US
• Bryan Catanzaro · Cupertino, US
• Gregory Diamos · San Jose, US
• Erich Konrad Elsen · Mountain View, US
• Ryan Prenger · Oakland, US
• Sanjeev Satheesh · Sunnyvale, US
• Shubhabrata Sengupta · Menlo Park, US
• Adam Coates · Mountain View, US
• Andrew Yan-Tak Ng · Palo Alto, US

Key dates

• Filing date: Jun 9, 2015
• Grant date: Jan 21, 2020
• Expiry date: Sep 13, 2036

Classification

• Technology area (CPC G): Physics
• CPC primary: G10L15/26
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

Abstract

Presented herein are embodiments of state-of-the-art speech recognition systems developed using end-to-end deep learning. In embodiments, the model architecture is significantly simpler than traditional speech systems, which rely on laboriously engineered processing pipelines; these traditional systems also tend to perform poorly when used in noisy environments. In contrast, embodiments of the system do not need hand-designed components to model background noise, reverberation, or speaker variation, but instead directly learn a function that is robust to such effects. A phoneme dictionary, nor even the concept of a “phoneme,” is needed. Embodiments include a well-optimized recurrent neural network (RNN) training system that can use multiple GPUs, as well as a set of novel data synthesis techniques that allows for a large amount of varied data for training to be efficiently obtained. Embodiments of the system can also handle challenging noisy environments better than widely used, state-of-the-art commercial speech systems.