Particle swarm optimization (PSO) is one of the most effective optimization methods for the black-box optimization problem. PSO involves a large number of particles that share information with each other to search for an optimal solution. The method in which a large number of search individuals cooperate to search for an optimal solution is called swarm intelligence optimization. In group intelligence optimization, the balance between exploration and exploitation is important. However, in PSO, it is unclear to what extent each parameter affects exploration and exploitation. Therefore, we proposed a deterministic PSO without probabilistic elements and analyzed the dynamics of PSO using the dynamical systems theory. Each particle in deterministic PSO has its motion determined by its eigenvalue. To make this motion clearer, a canonical deterministic PSO on a regularized phase space was proposed. The results of these analyses clarified what is attributed to the parameters for exploration and exploitation, i.e., global and local search capabilities. On the basis of this fact, we proposed a nonlinear map optimization (NMO) with improved local search capability. In this paper, we present the background of our proposal and consider the solution-search capability of NMO.

With the progress of biometric authentication technology, its accuracy is becoming more sophisticated. For this reason, a huge amount of biometric data is required for accuracy evaluation, and the collection cost is an issue for biometric authentication vendors. In order to reduce the collection cost, we have been researching and developing biometric accuracy evaluation methods using extreme value statistical models. These methods have been recognized and international standardization has begun. In this paper, we will explain the background of these studies, the new accuracy evaluation methods utilizing the developed extreme value statistics, the points and effects of applying extreme value theory, and future standardization trends.

Machine learning technologies such as deep neural networks (DNNs) (hereafter referred to as “AI”) demonstrate remarkable performance in various tasks such as image recognition, and the implementation of AI in society is expected to be further accelerated. At this time, it is important to consider security measures against attack methods such as inducing malfunctions and leaking privacy information on AI, in order to implement AI in safety or security applications such as those of autonomous driving vehicles or surveillance cameras. In addition, AI models must be protected, because trained AI models are important intellectual property. On the other hand, the inference process must be executed on embedded devices (hereinafter referred to as “edge AI”) in applications such as surveillance cameras and autonomous driving vehicles, which require privacy protection and processing without delay. In these cases, it is necessary to ensure security by assuming scenarios where an attacker can physically access the edge AI, and the research field of hardware security for edge AI has been activated since around 2017. In this paper, we review the threats of AI security and hardware security for edge AI and introduce recent research topics and countermeasures.

In general modulations, symbols are designed in two-dimensional Euclidean space, which is equivalent to the I-Q signal space, based on the quadrature modulation. The typical modulation schemes are phase shift keying (PSK) and quadrature amplitude modulation (QAM). Their symbols are designed to maximum the minimum Euclidean distance between the symbols under the condition of the constant average signal power. In this article, I explain the symbol design schemes in K-dimensional Euclidean space. In particular, I show three designs for the symbols on the K-1-sphere. It is possible to make the minimum distance between symbols longer than that of quadrature PSK (QPSK) and reduce E_b/N₀ required to achieve a BER. I explain the availability of hypersphere modulation in wireless communication systems.

One of the efforts to preserve a sound environment is regulation by based on the noise level. The noise level is a sensory quantity in which the loudness of the sound is weighted using the characteristics of hearing, but the pitch and/or timbre cannot be evaluated. As the movement to improve the quality of life becomes more active, activities leading to the evolution of the sound environment to a better one will become more vigorons in the future. Therefore, the pitch and timbre are also important factors in the evaluation of environmental sounds. In frequency analysis, the pitch and timbre characteristics can be investigated using the spectrum, but there is no index that represents the waveform characteristics, and it cannot be quantified or evaluated in the same way as the noise level. The authors propose to consider the pitch and timbre as the complexity of the waveform by measuring the complexity as the strength of self-similarity taking advantage of the fact that the fractal dimension can be used for the evaluation. Since the fractal dimension of the waveform can be expressed by one real number, it can be numerically evaluated similarly to the noise level. Previous studies have revealed that complexity can be discriminated by hearing. It is also known that the fractal dimension differs depending on the regional characteristics such as airport and railway lines. In this paper, we introduce an outline of the noise regulation law and its operation, and describe the background focusing on the complexity of waveforms, the method of evaluating complexity using the fractal dimension, analysis examples, and prospects.

It is believed that the use of RF signals in seawater is not possible owing to large attenuation except LF bands. Recently, the use of RF signals in underwater communications has been viewed as one of expected solutions towards realizing various applications in seawater. In this paper, results of experiments on radio communications conducted in shallow seawater are shown.