AI Times, a Korean AI publication, has published Managing Partner Yongduck Kim's column on AI patents.
Yongduck Kim, a patent attorney, focuses on patents in cutting-edge technology fields such as artificial intelligence (AI), smart factories, and blockchain, and has experience performing intellectual property (IP) work for various technology companies such as Samsung Electronics, LG Electronics, and Xiaomi. In addition, he has written books such as “AI Patent Examination Practice Guide” and “AI Patent Strategy” and continues to share his expert knowledge on AI technology and patent strategy.
The column examines machine learning for identifying transitions in continuous audio streams, including boundaries between music, advertisements, and presenters' speech.
The analysis centers on U.S. Patent No. 11,935,520, 'Identifying shifts in audio content via machine learning.' The described system combines overlapping three-second samples, MFCC feature extraction, CNN classification, analysis of relationships between adjacent samples, and HMM post-processing.
This column goes beyond just a technology introduction to shed light on what AI technology that detects the ‘bounds’ of continuous data can mean from a patent perspective. It analyzes the strategic value of system design combined with overlapping sampling and model interaction and probabilistic post-processing structure in terms of inventive step and scope of rights, and provides direction for AI patent in media, streaming, and content analysis in the future.
The full article can be found at the link below.
This article reflects the information available when it was published. Contact us to discuss your circumstances.