Ki Hun Hong

Title: Ki Hun Hong: Advancing Deepfake Voice and Video Detection

Introduction

Ki Hun Hong is an inventor based in Seoul, South Korea. His patent work addresses methods for detecting deepfakes in voice and video, with a focus on analyzing synthetic speech and inconsistencies between a speaker’s face and voice. He is associated with Soongsil University and the Electronics and Telecommunications Research Institute. His listed portfolio includes 3 patents.

Latest Patents

**Method and Apparatus for Learning Deepfake Voice Detection Model.** This invention describes training a model to detect fake speech. It resynthesizes a real voice sample with multiple vocoders to create fake samples, organizes real and fake samples into mini-batches, and uses supervised contrastive learning to obtain a vocoder signature for the detection model.

**Method and Apparatus for Detecting Voice-Based Deepfake Video.** This invention describes assessing whether a video of a speaking person is a deepfake. The method analyzes the direction of the person’s face over time in both the video imagery and voice, then compares the results to determine whether the video is manipulated.

Career Highlights

Hong’s listed affiliations include Soongsil University and the Electronics and Telecommunications Research Institute. His patent work reflects research into machine-learning approaches for identifying synthetic media.

Collaborations

His listed coworkers include Sou Hwan Jung and Kwi Hoon Kim.

Conclusion

Ki Hun Hong’s inventions focus on detecting deepfake speech and video by examining vocoder characteristics and comparing facial and voice-based signals.