臉部年齡辨識系統應用於人與機器人互動

Abstract

電腦科技和資訊科學的發展日新月異，在這樣的趨勢之下，智慧型機器人的重要性逐漸被應用於各個層面，包括自動化工業、軍事國防、保全巡邏、居家照護、教育娛樂等。而以人為核心的社會環境當中，人與機器人的互動關係便成為相當受到重視的一環，因此我們致力於開發各種人類資訊的偵測系統落實於電腦或機器人上，經由擷取人類的各種靜態與動態資訊，使得電腦或機器人有良好的先前知識來達到人機互動的目的，例如一個先進的健康照護系統，可以根據患者年齡自動配置適當的虛擬護士，隨時監控病情給予照顧，同時帶來互動的樂趣；另一方面藉由年齡資訊的擷取，我們能提升保全監控的應用面，例如管制未成年人從事違法行為等以降低人力資源的支出。 本論文主題為開發人類臉部年齡辨識系統，並應用於人與機器互動。一開始先說明年齡辨識的應用範圍以及重要性，接著針對人臉特徵的選擇進行比較與測試，在本論文中我們決定選用全相貌影像作為判斷年齡的特徵依據，並且將年齡分成共七個群組，每十歲為一組，大於60歲則皆屬於同一群組。我們採用FG-NET和MORPH這兩個西方人臉資料庫，以支持向量機做學習，最後結合F-measure的分數以加權的方式決定辨識出的所屬年齡群組。本系統所得到的辨識結果優於人的肉眼主觀判斷，將此一結果以即時辨識的方式和電腦做互動，藉由人機介面的設計讓使用者簡單操作，達到人機互動的效果。 本論文所提出的演算法與程式架構皆於Windows 7作業系統上以C++程式語言開發，且依據辨識結果，將所有軟體硬體整合，預期應用在本實驗室自行開發的智慧型機器人上。

Since computer technology and information science change with each passing day, intelligent robots become more important in a variety of fields, such as industrial automation, military defense, security guard, in-home nurse, education and entertainment and so on. For this human-centered society, the interaction between human and robot is regarded as a significant part of the technological environment. Therefore, many scientists and researchers dedicated their time to develop all kinds of human information estimation systems so as to implement on computers and robots. Through the detection of static and dynamic human information, the purpose of human-machine interaction could be achieved according to the information. There are many modern applications require the function of age estimation such as security control and surveillance monitoring, health care system and so on. In this study, we propose a method to classify human age using appearance images and apply it to the human-robot interactions. We first confirm that facial features based on craniology are not discriminative under the condition of seven age-groups classification. Next, our system is designed to have two stages. One is image preprocess stage; faces are detected and preprocessed. Our image database is from FG-NET and MORPH databases so that we have high degree of complexity in training dataset. Then images are trained by support vector machines (SVM). To have higher recognition rate, we train RBF (radial basis function) and linear kernel models at the same time, and decide the final results by F-measure based weighting policy. We also compare the age-group classification results with subjective questionnaires, and it demonstrates that the proposed system has better performance than human’s subjective estimation. For the purpose of human-machine interaction, we design a simple user interface to perform online age-group classification. The system can be applied on any computer or robot as long as it has a camera sensor. All the algorithms and software programs proposed in this thesis are implemented with C++ programming language in Windows 7 platform, and the integrated development environments are Microsoft Visual Studio.