具使用者回饋機制及情境感知之即時每日圖文摘要應用

Abstract

隨著各種不同的穿戴式裝置的問世與發展，穿戴式裝置逐漸地受到普羅大眾所重視而開始普及。大多數的穿戴式裝置收集使用者的資料，並根據收集到的使用者資料，做某些特定的事情，例如運動手環，會收集使用者的心跳資訊，以及移動步數…等，然後提供健康管理方面的分析。而由於穿戴式裝置越來越被一般人所接受，使用者可能攜帶不止一種的穿戴式裝置。可以預期的是，我們可以從個別使用者的不同穿戴式裝置中，收集到相當大量的資訊。這些穿戴式裝置，如果只是分別進行各自特定領域的分析，會相當的可惜。因此，本文提出一套系統並製作出一套應用，使得不同的穿戴式裝置所收集的資訊，得以整合，成為自動日記產生器，判別使用者所進行的活動，以及使用者感興趣的圖片的重要資訊。 此應用由穿戴式裝置、行動裝置組成，旨在收集使用者日常生活訊息並自動生成一個每日圖文活動摘要。使用者將穿戴攝影裝置週期性拍攝日常生活，接著行動裝置將自動進行活動辨識以及篩選出具有代表性的生活照，並呈現在手機的畫面上供使用者檢視。此外，使用者可對自動產生的內容進行修改，此應用會針對使用者修改的內容，對圖文預測的模型進行修正，藉此回授機制，使得此應用更易於使用。

With the appearance and development of various kinds of the wearable device, the wearable device has been placed a high value by the general public gradually. Most of the wearable devices collect data from the user, doing some work or analysis in specific area according to the data. For example, sport wristband collects the user data like heartbeat, number of steps … etc., providing the analysis about health care domain according to the data. As the wearable device being accepted by more and more people, more than one wearable device might be brought by a user. It is expectable that more and more information about a user could be retrieved from different kinds of wearable device they brought. It would be a pity that the collected data from those different kinds of wearable devices are only doing works in their specific domain. As a result, we proposed a system and developed an auto diary generator application, making the data from different wearable devices become important information for the activity prediction and image selection in our application. The application is composed of a wearable device and a mobile device, aiming at collecting the information of the user’s daily life and generating a diary-like summary in real-time automatically. The user wears the wearable camera, taking pictures once in a while. The mobile device will recognize the activity and select some representative pictures automatically, showing a diary-like summary in the user interface of the mobile device. In addition, users can modify their own auto-generated diary if they want to. The application will conduct adjustment on the classification model according to the modification of the user. With the feedback mechanism, the application is easy to use.