Keiji Shimada

Appearance-based smile intensity estimation by cascaded support vector machines

Facial expression recognition is one of the most challenging research area in the image recognition field and has been studied actively for a long time. Especially, we think that smile is important facial expression to communicate well between human beings and also between human and machines. Therefore, if we can detect smile and also estimate its intensity at low calculation cost and high accuracy, it will raise the possibility of inviting many new applications in the future. In this paper, we focus on smile in facial expressions and study feature extraction methods to detect a smile and estimate its intensity only by facial appearance information (Facial parts detection, not required). We use Local Intensity Histogram (LIH), Center-Symmetric Local Binary Pattern (CS-LBP) or features concatenated LIH and CS-LBP to train Support Vector Machine (SVM) for smile detection. Moreover, we construct SVM smile detector as a cascaded structure both to keep the performance and reduce the calculation cost, and estimate the smile intensity by posterior probability. As a consequence, we achieved both low calculation cost and high performance with practical images and we also implemented the proposed methods to the PC demonstration system.

Fast and Robust Smile Intensity Estimation by Cascaded Support Vector Machines

 Abstract—Facial expression recognition is one of the most challenging research areas in the image recognition field and has been studied actively for a long time. But it has not achieved enough performance under the practical environment yet. Especially, smile is the most important facial expression used to communicate well between human beings and also between human and machines. Therefore, if we can detect smile and also estimate its intensity at low calculation cost and high accuracy, it will raise the possibility of inviting many new applications in the future. In this paper, we focus on smile in facial expressions and study feature extraction methods to detect a smile and estimate its intensity only by facial appearance information (Facial parts detection, not required). We use Local Intensity Histogram (LIH), Center-Symmetric Local Binary Pattern (CS-LBP) or features concatenated LIH and CS-LBP to train Support Vector Machine (SVM) for smile detection. Moreover, we construct SVM smile detector as a cascaded structure both to keep the performance and reduce the calculation cost, and estimate the smile intensity by posterior probability. As a consequence, we confirmed that our proposed method provided the comparable performance with the existing method, and it also achieved both low calculation cost and high performance even with the practical database. The visual information plays a very important role in our everyday life. Especially, in regard to communication between human beings, we can come to understand deeply and smoothly each other to pay attention to behaviors and facial expressions as well as languages. Facial expression analysis has been approached by several research fields, for example in psychology [1], brain science, etc. In engineering [2] too, many researchers have tried to analyze, estimate and understand facial expressions and human emotions by face images, by voice signals, by bio-signals, etc. for a long time. But, it is still difficult to recognize facial expressions only by face images automatically, because there are many problems such as inconsistencies in individuals, lack of criterion to judge facial expressions, disparities between simulation data and practical data and a mismatch between the expressions and the emotions. Therefore, there is no critical solution to recognize those by computer automatically under the practical environment and active research is still very much in progress. In particular, smile (In a wide sense, facial expressions, which are observed when human beings derive pleasure) is one of the most important facial expression used to …

The Assessment of Driver's Arousal States from the Classification of Eye-Blink Patterns

To realize the real-time assessment of drivers arousal states, we propose the assessment method based on the analysis of eye-blink characteristics form image sequences. The drivers arousal level while driving is not monotonous falling from high to low. We proposed the two-dimensional arousal states transition model which was taken into account the fact that a driver usually held out against sleepiness. The eye-blink pattern categories were classified from image sequence using HMM (Hidden Markov Model), then the drivers arousal states were finally assessed using HMM by histogram distribution of those typical eye-blink categories. The arousal assessment results are also verified against the rating results by trained raters.

The home networking system based on IEEE1394 and Ethernet technologies

We propose the home networking system which includes various network technologies. Our home networking system consists of low-power IEEE1394 and Ethernet wall-plate, multi-core POF & UTP Cat.5 combination cable, IEEE1394-Ethernet bridge router and network control application based on Web server technologies. The characteristics of our home networking system are user friendly operation, seamless control, low-power and low-cost.