The emergence of internet video streaming and home digital recording systems has resulted in a plethora of entertainment viewing options for the consumer. Despite this, access to music videos and individual songs, especially in broadcast television, are dictated by the content provider. While tools for audio analysis of music video content exist, an alternative which also considers visual content is valuable for providing an enjoyable and convenient interface with the consumer. This technology is an algorithm for automatically segmenting music videos from a longer entertainment block and categorizing them based on audio and visual content. Using this technique, music video summaries condensing the information from the videos can then be produced for users to browse. This technology provides a new, intuitive method of presenting music video content to users.
This technology performs automatic content analysis in order to summarize music videos embedded in entertainment segments which may also contain non-music video content. The algorithm performs analysis on high level information including, but not limited to: videotext, face detection, camera changes, color histogram, audio features, and lyrical information. Using pattern recognition, the beginning and end of individual music videos can be identified accurately. Utilizing the content elements mentioned, a music video summary is produced which condenses the music video and categorizes it by a variety of elements such as visual themes, or song structure.
A working prototype of the algorithm has been produced and tested with user surveys indicating satisfaction with a music video summary interface.
Patent Issued (US 7,599,554)
Tech Ventures Reference: IR CU13213