The evolution of CMOS image sensors and the prospects utilizing advanced imaging technologies promise to improve our quality of life. CMOS image sensors now dominate the market of digital cameras with the advent of parallel ADCs and back-illuminated technology. Moreover stacked CMOS image sensors continue to enhance functionality and user experience in mobile devices. We introduce the evolution of image sensor architecture for accelerating drastic performance improvements and for integrating edge computing connected to the sensor layer through stacking technologies. Furthermore, the fine pitch connection between the pixel and logic layers makes pixel parallel circuit architecture practical for the next stage of evolution.

The installation of trillion sensors will realize a connected society able to provide never-before-seen public services, such as disaster prediction by exploiting massive data via machine-learning. However, large-scale deployments of battery-driven sensors will cause unacceptably huge maintenance costs because of the finite amount of energy in the batteries. Harvesting energy from natural phenomena such as light, heat, vibration, chemical reaction, and radio waves is a promising solution for overcoming this limitation and enabling perpetual operations of sensors. Since energy harvesting (EH) is inherently unstable and uncontrollable, the design of wireless communication with EH is different from that with a stable power source. In this paper, the basics of EH technologies are introduced and a brief survey of recent results on communications design with EH technologies for different network topologies is provided.

Phase retrieval is a technology of reconstructing a complex-valued signal from its magnitude. Since the information missing in magnitude is the complex argument, or phase, the reconstruction of a complex-valued signal is achieved by retrieving its phase. As has been studied in the field of optics, the mainstream formulations and methods of phase retrieval are based on situations arising from optical measurement. At the same time, phase retrieval has been studied with different settings and motivations in the field of acoustical signal processing, where some specific techniques have been considered. In this paper, the historical background and general phase retrieval are introduced in the first half. Then, in the latter half, phase retrieval in acoustical signal processing based on the short-time Fourier transform is reviewed.

Machine learning is an assembly of technologies that can learn an appropriate way of processing data from data themselves, rather than by explicitly specified by humans. It is a new software paradigm that brings about much progress in many application domains, but it also has many uncharted areas. Recently, it is found that there are types of security flaw specific to machine learning, which are now being actively studied. In this article, we illustrate the research trends of machine learning security and also show a perspective to realize secure applications of artificial intelligence (AI).