Overview of research

Currently, multimedia data accounts for up to 70-80% of Internet traffic. Also, visual contents are increasingly being captured and generated by IoT (e.g. autonomous cars, drones, robots, smart cameras, etc.). This fact results in big challenges for both providers and users in accessing, analyzing, and disseminating the huge amount of multimedia contents today. In addition, there is a convergent infrastructure for entertainment and communications services in our daily life. Currently, we focus on the following topics.

Multimodal big data processing for Smart society

In our smart society, the physical world and digital world are getting closer and closer. Humans is increasingly working with robots, virtual objects, virtual humans. In this context, it is important to process and analyze the multimodal big data provided by IoT devices to enhance human life. Currently, we are processing new multimedia of smart camera and virtual reality, and then extracting the semantics of the contents.

AI, deep learning, quantum machine learning

On the other hand, devices like robots and autonomous cars need to understand the surrounding situations like a human. Also, the interactions between human and the digital world should be understood and improved for good Quality of Experience (QoE). We are investigating new tools of classical machine learning, deep learning, and quantum machine learning for these purposes.

Multimedia and AI over networks

In a recent study, Cisco predicts that 90 percent of the bits carried on the Internet will be video traffic in the near future. In this research, our purpose is to investigate adaptative and low-delay solutions to network fluctuations. Further, the impacts of transport (e.g. low resolution, low frame rate, delay) and adversarial attacks on AI performance will be studied.

IoT System Supporting Senior Citizen

(Co-work with UoA's IoT group)

With a fast aging society, elderly people living alone are increasing in number. The elderly people have various needs, such as assist for living and health, monitoring of safety, and relief from uneasiness and loneliness. In this research, a smart town which utilized IoT and M2M technology is built for elderly people's support. Meanwhile, elderly people's life situations are recognized automatically and shared with the persons concerned, such as family and neighbors. Besides using wired internet, Wi-Fi mesh networks will be employed and optimized for sending such situation information including streaming information.

e-Health System

(Co-work with the Cluster for Medical Informatics and Engineering)

Diagnosis and prevention of depressive disorders at any scale have been attracting considerable attention of the public healthcare in Japan because depression is one of the most rapidly pervasive mental disorders in the country. A major issue that hinders the feasibility of depression screening for its prevention is the availability of some simple and cost-effective methods for depression detection and monitoring. In this research, we investigate a computerized tool for depression detection. The tool utilizes the theory of chaos and systems complexity to extract robust dynamically statistical features of physiological signals provided by the low-cost technology of photoplethysmography. Prototype of a mobile communication system for the proposed automated depression detection is developed to illustrate a potential e-mental health system for automated screening and detection of depression.

Demonstrations

Link to some previous research