運用感測器於閱讀專注度偵測

Abstract

隨著感測元件技術與行動運算能力的進步，許許多多的行動穿戴裝置孕育而生。現有的行動穿戴裝置，主要聚焦於人體姿態、人體運動的議題上。我們留意到近年來，開始有些除了探測人體運動外的議題，這些議題是有關於人類注意力方面的議題，像是學生閱讀注意力觀測。但大多數的閱讀注意力研究以電子書為主，且都需要於人體穿戴感測元件或定點偵測，較少考慮以閱讀實體書籍為基礎的研究。 有鑑於此，本研究運用感測元件與實體書籍結合，以間接式感測的思維角度為出發，切入人類閱讀行為，反向推導書本與人之間的互動關係，使人體不需再為穿戴感測元件感到不便及受定點拘束，幫助人們了解閱讀過程的專心程度，進而達到自我檢視的功效。 我們提出一套名為 RIM (Read Identification Method)的演算法，利用 多種模式的篩選過濾，達到判別閱讀專心的成效。並透過特徵抽取以機器學習KNN 方法進行判別閱讀專心程度，以此做為比較的對照組。本研究提出的方法原型實作於Ardunio平台上，並以真實受測者收集的資料，來驗證提出的方法實際效率，實驗結果顯示，我們提出的方法，能夠有效率地辨識閱讀過程的專心程度，其平均正確率高達95%。

With the sensor technology and mobile computing ability to progress,many mobile wearable device was born. We noticed that in recent years, in addition to detect human motion to start some issues,these issues are related to human beings attention to issues like the students reading attention observation. But most research attention to reading e-books, and the need for human wearable sensing element or point detection,on the basis of reading physical books considered less. This research uses the sensing element and the physical book binding,with indirect sensing thinking perspective,the breakthrough of human reading behavior,interaction between reverse books and people, so that the body does not need to wear the sensing element to inconvenience and point restraint,help people learn to concentrate on reading the process level,and achieve the self view effect. We present a set of named RIM (Read Identification Method) algorithm,using the filtering of many patterns to determine reading attentively results. And through the feature extraction by KNN machine learning method for identification of reading concentration degree,as compared to the control group. This study proposes a method of implementing a prototype on Ardunio platform, and based on real subjects collected data to verify the proposed method of actual efficiency,the experimental results show that our proposed method can efficiently identify,the process of reading the concentration degree,the average correct rate of 95%.