Nikhil Singh

Title: Nikhil Singh and Language-Invariant Audiovisual Representations.

Introduction

Nikhil Singh is an inventor based in Los Gatos, California, and works at Netflix Inc. His patent work addresses machine-learning methods for representing scenes in long-form video despite differences in spoken language.

Latest Patents

Singh is associated with 1 patent: “Methods and Systems for Learning Language-Invariant Audiovisual Representations.” The invention describes training a machine-learning model with video clips and corresponding audio spectrograms, including dubbed audio in primary and secondary languages. This approach teaches the model to de-emphasize speech audio when generating audiovisual representations of similar scenes, helping it recognize scene content across language differences.

Career Highlights

The disclosed invention focuses on preparing training data that pairs video clips with audio spectrograms in different languages. By using this data to train the model, the system aims to produce representations that are less dependent on the language being spoken.

Collaborations

The provided information names Iroro Orife and Chih-Wei Wu as Singh’s coworkers. Their specific roles in the patent or related work are not stated.

Conclusion

Nikhil Singh’s patent describes a machine-learning approach to audiovisual representation that reduces the influence of spoken-language differences in long-form video.