PUB - Publikationen an der Universität Bielefeld

This paper presents a diary system which reduces the time to keep and maintain a diary by assisting the user with automatic suggestions for possible events and activities. The system uses a smartphone app reading the internal sensors to detect the current situation. It interrupts the user when unsure in prediction of the current situation. No external hardware is used to assure a broad audience of later end-users. The overall goal is to create a multimedia enriched digital personal diary. A study has been conducted where 16 participants collected data over two weeks from the built-in sensors of an ordinary smartphone. The data has been annotated by the users with labels of daily activities. We applied different classification algorithms to test the feasibility to detect these activity labels from the collected sensor data only. The experiments show that a perfect classification of activities of daily living is not possible but a system like this is usable to assist the journaling by suggestions and semi-automatic logging. The digital diary keeping process should take less time and is more versatile than an old-fashioned handwritten diary. A diary like this can be used to revive past events and help people with dementia to remember the past.