CONF Korchagin_UCM_2009/IDIAP Memoirs of Togetherness from Audio Logs Korchagin, Danil confidence estimation pattern matching time-frequency analysis EXTERNAL https://publications.idiap.ch/attachments/papers/2009/Korchagin_UCM_2009.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Korchagin_Idiap-RR-36-2009 Related documents Proceedings International ICST Conference on User Centric Media Venice, Italy 2009 P.O. Box 592, CH-1920 Martigny, Switzerland December 2009 In this paper, we propose a new concept how tempo-social information about moments of togetherness within a social group of people can be retrieved in the palm of the hand from social context. The social context is digitised by audio logging of the same user centric device such as mobile phone. Being asynchronously driven it allows automatically logging social events with involved parties and thus helps to feel at home anywhere anytime and to nurture user to group relationships. The core of the algorithm is based on perceptual time-frequency analysis via confidence estimate of dynamic cepstral pattern matching between audio logs of people within a social group. The results show robust retrieval and surpass the performance of cross correlation while keeping lower system requirements. REPORT Korchagin_Idiap-RR-36-2009/IDIAP Memoirs of Togetherness from Audio Logs Korchagin, Danil EXTERNAL https://publications.idiap.ch/attachments/reports/2009/Korchagin_Idiap-RR-36-2009.pdf PUBLIC Idiap-RR-36-2009 2009 Idiap P.O. Box 592, CH-1920 Martigny, Switzerland December 2009 In this paper, we propose a new concept how tempo-social information about moments of togetherness within a social group of people can be retrieved in the palm of the hand from social context. The social context is digitised by audio logging of the same user centric device such as mobile phone. Being asynchronously driven it allows automatically logging social events with involved parties and thus helps to feel at home anywhere anytime and to nurture user to group relationships. The core of the algorithm is based on perceptual time-frequency analysis via confidence estimate of dynamic cepstral pattern matching between audio logs of people within a social group. The results show robust retrieval and surpass the performance of cross correlation while keeping lower system requirements.