Accurate repeat finding and object skipping using fingerprints

This paper introduces a novel and very accurate segmentation algorithm. It is very efficient and consumes less than 10% of CPU on a simple desktop PC to segment a stream in real-time. It operates on an audio stream, or on the audio portion of a audio-visual stream. It is very accurate: it accurately detects the positions and durations of objects on an over-the-air broadcast television signal, and songs on both FM and internet radio stations (as checked against labeled ground truth streams). The algorithm does not require any prior information or training. We detail the system design and present results of segmenting broadcast streams. Keywords Repeat finding, Segmentation

• Type: Conference
• Year: 2005
• Where: MM
• Authors: Cormac Herley
• Keywords: Accurate Segmentation Algorithm, MM 2005, Over-the-air Broadcast Television, Simple Desktop Pc